Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
DEPARTMENT OF PRODUCTION
ENGINEERING
e-Lecture Notes on
PRECISION ENGINEERING (PE)
COURSE CODE: BMS 408
MODULE(I,II,III,IV)
th
8 Semester B. Tech
Production Engineering
Module-I
Manufacturing is the cornerstone of many industrial activities and significantly contributes toward the
economic growth of a nation. Generally, the higher the level of manufacturing activity in a country, the
better is the standard of living of its citizens. Manufacturing is the process of making large quantities of
products by effectively utilizing the raw materials. It is a multidisciplinary design activity simply involving
the synergistic integration of production and mechatronics engineering. The products vary greatly from
application to application and are prepared through various processes. It encompasses the design and
production of goods and systems, using various production principles, methodologies and techniques. The
concept is hierarchical in nature in the sense that it inherits a cascade behaviour in which the
manufactured product itself can be used to make other products or items. The manufacturing process may
produce discrete or continuous products. In general, discrete products mean individual parts or pieces
such as nails, gears, steel balls, beverage cans, and engine blocks, for example.
Conversely, examples of continuous products are spools of wires, hoses, metal sheets, plastic sheets,
tubes, and pipes. Continuous products may be cut into individual pieces and become discrete parts. The
scope of manufacturing technology includes the following broad topics:
Controls
Dimensional metrology and surface metrology
Instrument/machine design
Interferometry
Materials and materials processing
Precision optics
Scanning microscopes
Semiconductor processing
Standards
Frequently used terms within the domain of precision and ultra-precision engineering are precision
processes, scaling, accuracy, resolution and repeatability. The precision process is a concept of design,
fabrication, and testing where variations in product parameters are caused by logical scientific
occurrences. Identification of these logical phenomena and strategically controlling them is very
fundamental to precision manufacturing. Scaling is a parameter that defines the ratio attributes with
respect to the prototype model. It is also considered as a fundamental attribute for predicting the
behaviour of structures and systems for analysis and synthesis of miniaturised systems. Accuracy defines
the quality of nearness to the true value. In the context of machine or production systems, accuracy is the
ability to move to a desired position. As an example, if the actual value is 1.123 units and it is recorded as
1.1 units, we are precise to the first decimal place but inaccurate by 0.023 units. Resolution is the fineness
of position precision that is attainable by a motion system. The smallest increment that is produced by a
servo system is the resolution. There are two types of resolutions, electrical and mechanical. With regard
to mechanical resolution, it is defined as the smallest increment that can be controlled by a motion system,
i.e., the minimum actual mechanical increment. One can note that mechanical resolution is significantly
coarser than that due to the involvement of friction, stiction, deflections, and so on. Repeatability is the
variation in measurements obtained when one person takes multiple measurements using the same
instruments and techniques. Repeatability is typically specified as the expected deviation, i.e., a
repeatability of 1 part in 10,000 or 1:10,000, for example.
The major problem of conventional laser microdrilling is that the process has a short focal depth. It is
known that this method typically achieves aspect ratios up to 100 in thick material, such as for a 15-μm hole
in 1.5-mm-thick foil, for instance. This problem can be overcome by utilizing a Bessel beam. Deep high
aspect ratio drilling is achieved due to the reason that the Bessel beam is nondiffracting and in practice
they do not spread out. In the case of deep high-aspect ratio laser drilling, a pseudo-Bessel beam is
generated using a pulsed laser. Some of the examples of microdrilling applications using a laser system
developed by ATLASER di Andrea Tappi are presented in Table 1.1. The application of lasers to
micromanufacturing has several advantages: noncontact processing, the capability of remote processing,
automation, no tool wear and the possibility of machining hard and brittle materials.
1.5 Nanotechnology
Nanoscale devices and equipment provide benefits in terms of an improvedgreener environment,
miniaturization, efficiency and resource consciousness.Nanotechnology has accelerated research and
development in many disciplines.However, a key obstacle to its development remains in the need for cost-
effective large-scale production methods. Nanotechnology has applications in many fieldsincluding
automotive, aerospace, household appliances, sporting goods,telecommunication equipment and medical
supplies.
The ongoing research that is being carried out all over the world is based on thestudy of conductive and
high-strength composites, energy storage and energyconversion devices, sensors for field emission
displays and radiation sources,hydrogen storage media and nanometer-sized semiconductor devices such
asprobes and interfacings. Nanotube-based design scenarios anticipate the design ofgears and bearings
and hence the development of machines at the molecular level.The team at the Paul Pascal Research
Center has been working on overcomingthe obstacle of the formation the carbon nanotubes and has
developed a processfor aligning them in the form of fibers and strips. Variorum structuralconstructions can
be formed through appropriate methods. Continuous sputteringof carbon atoms from the nanotubes lead to
a dimensional change, which facilitates surface reconstruction with annealing. An X-like junction with
diverseangles between the branches can be formed. Under careful irradiation one of thebranches of the X-
junction can be removed, thereby creating Y- and T-likejunctions. This new class of carbon junctions
exhibits an intrinsic non-lineartransport behavior, depending mainly on its pure geometrical configuration
and onthe kind of topological defects.Calculation and measurement of characteristic curves, like current
versus voltageof different sets of Y junctions, show robust rectification properties, giving rise tothe
possibility of using these junctions as nanoscale three-point transistors. Advanced computational
techniques, including large-scaleparallelisable molecular dynamic simulations of the growth mechanism
and first principlescalculations of the electronic structure, are being applied to model theself-assembly and
the electronic properties of nanostructures.Based on computational results, Han et al. at NASA Ames
Research Center havesuggested that nanotube-based gear (Fig.1.1(b)) can be made and operated and
thegears can work well if the temperature is lower than 600-1000 K and the rotationalenergy is less than
the teeth tilting energy at 20°.
Fig. 1.1. (a) A carbon nanotube (b) A carbon nanotube-based gear
2 Principles of MEMS
A considerable amount of research is being carried out concerning the design and development of existing
systems that reach down into micro- and nanometer scale levels. A technology that considers microscale
sensors, actuators, valves, gears, and mirrors embedded in semiconductor chips is referred to as
microelectromechanical systems, MEMS in short. In essence, MEMS are small, integrated devices that
combine electronics, electrical as well as mechanical elements (Fig. 2.1). The size is in the order of a
micrometer level. MEMS design technology is an extended form of traditional fabrication techniques used
for IC (Integrated Circuit) manufacturing. MEMS add passive elements such as capacitors and inductors
including mechanical elements such as springs, gears, beams, flexures, diaphragms, etc. MEMS are thus
the integration of these elements on a single substrate (wafer) developed through more advanced
microfabrication and micromachining technology. While the ICs are fabricated by the use of IC process, the
mechanical micro components are fabricated using micromachining processes. This process helps in
etching away the parts of the selected portions of the wafer. The process can also add new structural
layers to form mechanical as well as electromechanical components. Thus, MEMS technology promises to
revolutionise many products by combing microfabrication-based microelectronics with micromachining
process sequences on silicon, making it possible for the realisationof a complete systems-on-a-chip (SoC).
The technology allows for the development of smart systems and products inheriting increased
computational capability, perception and control attributes. Smart systems can lead to expand the scope of
possible solutions to diagnostics for target applications. It has been mentioned that microelectronic
integrated circuits can be thought of as the brains of a system while MEMS augments the decision-making
capability with eyes and arms, to allow microsystems to senseand control the environment. MEMS devices
are manufactured by the use of batch fabrication techniquessimilar to those used for IC. Therefore,
unparalleled levels of superiority,sophistication, functionality, reliability and availability can be achieved on
a smallsilicon chip at a relatively low cost. Two important microsystems are microsensorsand
microactuators.Sensors gather information from the environment. The commonly usedtransduction
principles are chemical, thermal, biological, optical, magnetic andmechanical phenomena. Accordingly,
there are various types of microsensors. Theintegrated electronics process the information derived from
the sensors. In manycases the decision-making logics are integrated into the devices. The decision
ismostly transmitted to the actuator in order to achieve moving, positioning,regulating, pumping or filtering
actions. In this way, the environment can becontrolled depending on the desired purpose. The study of
MEMS accommodatesthe topics listed below. These principles are presented in this chapter.
Fabrication processes
Mechanical sensors and actuators
Thermal MEMS
Magnetic MEMS
Micro-opto-electromechanical systems (MOEMS)
Fig. 2.1. Microelectromechanical systems
Electrostatic drive
Magnetic drive
Piezoelectric drive
Electrothermal drive
Each driving principle has specific advantages and disadvantages with respect to deflection range,
required force, environmental durability, and most importantly the response time. Furthermore, the
required power supply mainly depends on the driving principles involved (Table 2.1).
Electrostatic drive is based on electrostatic forces between the microelectrodes. When an external voltage
is applied between the electrodes, a potential energy is stored which enables the actuation. The
electrostatic forces act perpendicular to the parallel electrode. Electromagnetic actuation is primarily a
current controlled process. The driving mechanism again requires currents of the order of several
hundreds of milliamps, and voltages in the range of less than one volt. Magnetic drive is an attractive
driving principle and very suitable for applications like dustfilled environments and in environments where
low driving voltages are acceptable or desired. Piezoelectric driving is based on the material properties of
crystals, ceramics, polymers, and liquid crystals. In a piezoelectric material, the internal dielectric
displacement is developed via an applied electric field and mechanical stress. Electrothermal devices use
electrically generated heat as an energy source for actuation. The electrothermal effects can be divided
into three different categories: shape memory alloys, electrothermal bimorphs, and thermopneumatic
actuators. Many electrostatic MEMS actuators that are investigated include micromotors, comb drive
actuators and microvalves. Accelerometers, ink jet printer heads, color projection displays, scanning
probe microscopes, pressure, temperature, chemical and vibration sensors, light reflectors, switches,
vehicle control, pacemakers and data storage devices are examples of high-end applications of MEMS
sensors and actuators.
Bulk micromachining
Surface micromachining
Micromolding
Bulk micromachining makes micromechanical devices by etching deeply intothe silicon wafer. It is a
subtractive process that involves the selective removal ofthe wafer materials to form the microstructure,
which may include cantilevers,holes, grooves, and membranes.The majority of currently used MEMS
processes involve bulk etching. In lightof newly introduced dry etching methods, which are compatible
withcomplementary metaloxidesemiconductors, it is unlikely that bulkmicromachining will decrease in
popularity in the near future (Kovacs 1998). Theavailable etching methods fall into three categories in
terms of the state of theetchant: wet, vapor, and plasma. The etching reactions rely on the oxidation
ofsilicon to form compounds that can be physically removed from the substrate(Kovacs, 1998). Conversely,
surface micromachining technology makes thinmicromechanical devices on the surface of a silicon
wafer.The surface micromachining sequences are as follows,
Wafer cleaning
Blanket n+ diffusion of Si substrate
Passivation layer formation
Opening up of the passivation layer for contacts
Stripping of resist in piranha
Removal of thin oxide through BHF etchant systems
Deposition of a base, spacer or sacrificial layer using phosphosilicate glass (PSG)
Densification at 950°C for 30-60 min in wet oxygen
Base window etching in BHF for anchors
Deposition of structural material deposition (e.g., poly-Si using CVD method at about 600°C, 100 Pa
and 125 sccm at about 150 Å/min)
Anneal of the poly-Si at 1050°C for 1 hour to reduce stress in the structure
Doping: in-situ, PSG sandwich and ion implantation
Release step, selective etching of spacer layer.
The micromolding process involves use of molds to define the deposition of thestructural layer. In this
case, the structural material is deposited only in those areasconstituting the microdevice structure. This is
apparently in contrast to both bulkand surface micromachining processes. Feature blanket deposition of
the structuralmaterial followed by etching to realise the final device geometry is done in onestep. Once the
structural layer deposition is over the mold is dissolved by using achemical etchant. Note that the etchant
does not corrugate the structural material.One of the most widely used micromolding processes is the
LIGA process. LIGAis a German acronym standing for lithographie, galvanoformung and abformung,or
lithography, electroplating, and molding. Photosensitive polyimides are mostlyused for fabricating plating
molds.
Fig. 2.2. Cantilever beam for the measurement of static and dynamic acceleration
Fig. 2.2 shows a typical cantilever sensor that can measure acceleration. Thebody of the sensor (proof
mass) could be about 0.5-0.7 micrograms. The proofmass moves in the X- and Y-axes. Polysilicon springs
suspend the MEMSstructure above the substrate facilitating the proof mass to move freely.
Acceleration causes deflection of the proof mass from its centre position. Therecould up to 32 sets of radial
fingers around the four sides of the square proof mass.The fingers (middle one), shown in the figure, are
positioned between two platesthat are fixed to the substrate. Each finger and pair of fixed plates
constitutes adifferential capacitor, and the deflection of the proof mass is determined bymeasuring the
differential capacitance. This sensing method has the ability ofsensing both dynamic acceleration such as
shock or vibration, as well as staticacceleration such as inclination or gravity.Many accelerometers employ
piezoelectric sensing techniques. Thin filmpiezoelectric materials such as lead zirconatetitanate (PZT) are
promisingmaterials for MEMS applications due to their high piezoelectric properties. Piezoelectricpolymers
are now being used for sensor applications. Piezoelectric polymericsensors offer the advantage of strains
without fatigue, low acoustic impedance andoperational flexibility. The PZT converts mechanical
disturbances to electricalsignals. The starting material for the front-side process (FSP) is a silicon
waferthat has silicon dioxide, lower metal electrode (Ti/Pt), and deposited PZT films.Industrial applications
of MEMS accelerometers include airbag releasemechanisms, machinery failure diagnostics, and
navigational systems.
2.4.3 Microphone
Acoustic MEMS are air-coupled, and can offer a wide range of applications suchas detection, analysis and
recognition of sound signals. The basic component is themicrophone, called micro-microphone or simply
MEMS microphone. A simpledefinition of a microphone is that it is an electromechanical acoustic
transducerthat transforms acoustical energy into electrical energy. It is an ultrasonicmicrosensor, which
takes advantage of miniaturisation and also consumes lowpower. When multiple microphones are arranged
in an array, the device is referredto as smart system as it can offer more reliable and intelligent operations.
Thebasic challenge that is encountered in designing the MEMS microphone is theformulation, design, and
implementation of signal processing circuitry that canadapt, control and utilise the signal in noisy
environments. MEMS microphonesare also very suitable for outdoor acoustic surveillance on robotic
vehicles, windnoise flow turbulence sensing, platform vibration sensing, and so on.
Fig. 2.3. Cross-sectional diagram of piezoelectric microphone; plan view
The ultrasonic sensor is based on the mechanical vibration of micro membraneor diaphragm realised in the
silicon platform. The diaphragm is a thin, circularmembrane held in tension and clamped at the edge. Many
designs use a flat freeplatethat is held in proximity to the back plate by electrostatic attraction. Thefree-
plate makes up a variable capacitor with the back plate. Its value changes during the vibration caused by
the sound signal. The deformation or deviation ofthe membrane (free-plate) from the normal values
depends on the amplitude of theincident pressure. Fig. 2.3 shows the schematic as well as a
micromachined SEM(Scanning Electron Microscope) picture of a MEMS microphone. The microphonecan
have a very low stray capacitance, is self-biasing, mass producible, arrayable,integrable with on-chip
electronics, structurally simple and extremely stable overtime in an ordinary environment. The typical
dynamic range is from 70 to 120 dBSPL and the sensitivity can be in the order of 0.2 mV/Pa over the
frequency range100-10 kHz.
2.4.4 Gyroscope
MEMS gyroscopes (Fig. 2.4) are typically designed to measure an angular rate ofrotation. A measurement
of the angle is useful in many applications. A verycommon application is the measurement of the orientation
or tilt of a vehiclerunning in a curved path. The MEMS gyroscope design introduces sophisticatedand
advanced control techniques that can lead to measure absolute angles. Somedesign is based on the
principle of measuring the angle of free vibration of asuspended mass with respect to the casing of the
gyroscope. Thegyroscope can accurately measure both the angle and angular rate for lowbandwidth
applications. The measurement of orientation, for instance, is veryuseful in the computer-controlled
steering of vehicles as well as for differentialbraking systems for skid control in automobiles. A typical
gyroscope consists of asingle mass, oscillating longitudinally with rotation induced lateral deflectionsbeing
sensed capacitively. The iMEMS ADXRS gyroscope from Analog Devices,Inc., integrates both an angular
rate sensor and signal processing electronics onto asingle piece of silicon. Mounted inside a 7x7x3 mm
BGA package, the gyroconsumes 5 mA at 5 V and delivers stable output in the presence of
mechanicalnoise up to 2000 g over a reasonably wide frequency range. A full mechanical andelectronic
self-test feature operates while the sensor is active. Mechanical sensorsinherit many drawbacks as follows,
Fig. 2.4. (a) Conceptual schematic of a torsional micromachined gyroscope with nonresonantdrive; (b)
Schematic diagram of a vibratory gyroscope
2.5.1 Thermometry
Rigorous thermal analyses and experiments that assist in designing MEMSstructures are in progress. Some
effort has been made on liquid-crystalthermometry of micromachined silicon arrays for DNA replication.
This work isbeing carried out at Perkin-Elmer Applied Biosystems. Replication is achievedthrough what is
known as the polymerase chain reaction (PCR) process. PCRrequires accurate cycling of the liquid sample
temperature with an operationalrange between 55 and 95oC. PCR that makes use of micromachined
structures isshown in the Fig. 2.5(a). The structure assures uniformity as far as temperatureand cycle
timing is concerned. It also utilises less reagent and samplevolumes. Thermal design requires
measurement of the temperature distribution inthe reacting liquid. The measurement is possible by
encapsulating the liquidcrystals suspended in the liquid. This in turn led to the measurement of
thetemperature uniformity and the time constant for about 20 vessels in amicromachined silicon array. Two
separate sets of crystals are used to imagetemperature variations near the two processing temperature
thresholds with aresolution of 0.1oC. While the thermometry technique described above is usefulfor
characterizingmicrofabricated PCR systems, it can also support thermaldesigns of a broad variety of
MEMS fluidic devices.
Fig. 2.5. (a) Liquid-crystal thermometry based micromachined vessel array for thermalprocessing of DNA
using polymerase chain reaction (PCR)
Thermally activated systems are mostly employed in the bio-analyticalmicrosystems or lab-on-a-chip (loac)
devices. Several key issues in existing andemerging bio-analytical microsystems are directly related to
thermal phenomena.Research and development on loac technology is essentially directed
towardminiaturisation and the integration of chemical and biochemical analysis tools formanipulation,
handling, processing and analysis of samples in a single integratedchip. In order to develop a highly
sensitive and responsive thermal device, the thermal mass of the transducer element is kept as low as
possible. This is only achieved by using thin film structures made by micromachining techniques.
2.5.4 Thermoactuators
Thermal actuation is another scenario with respect to its counterparts includingelectrostatic and
piezoelectric types. Thermal actuation is based on electrothermalenergy density transformation which is
given by, E=V2/L2, where V is theapplied voltage, is resistivity and L is the effective length of the
actuatorelement. Microactuators based on electrothermal principles can offer significantenergy density
compared to electrostatic microactuators. It is worth mentioningthat for some configurations,
electrothermal actuation provides 1 to 2 orders ofmagnitude higher energy density than piezoelectric
actuation, and 4 orders ofmagnitude higher energy density than electrostatic transduction.Several thermal
microactuator geometries have been investigated, including Ubeamand V-beam geometries,
respectively.Piezoelectric microactuators usually provide a fraction of the energy density of electrothermal
elements.
An example of a bi-directional magnetic actuator used for optical scanningapplications can be given. It is
composed of a silicon cantilever beam and anelectromagnet. At the tip of the cantilever beam, a permanent
magnet array is electroplated in order to achieve the bi-directional actuation. Below the cantileverbeam,
the permanent magnet array is placed along the axis of the electromagnet. For a large bi-directional
deflection and dynamic scanning capability,Cho has designed an optical scanner supported by two
serpentine torsion bars. Aschematic diagram of the scanner is illustrated in Fig. 2.7. The scanner is
designedto have a silicon mirror supported by two serpentine torsion bars (Fig. 2.8),carrying a permanent
magnet made by bumper filling at its tip on the oppositeside.
Fig. 2.7. Cantilever beam magnetic microactuator. (a) Schematic view and (b) Anotherview
Fig. 2.8. Magnetically driven optical scanner
Son and Lal reported a novel magnetic actuatorthat allows remote magnetic actuation with piezoresistive
feedback formicrosurgery applications. The actuator consists of an electromagnet, aferromagnetic mass,
and a cantilever. Conventional actuator fabricationlithography combined with electroplating is required to
make high aspect ratiostructures. However, high aspect ratio columns with a suspension of
ferromagneticnanoparticles and epoxy using magnetic extrusion were fabricated. The fabricationprocess
did not require lithography. The remote magnetic actuator with feedbackconsists of two basic units. They
are the actuation unit and feedback unit. Theferromagnetic mass is first placed on the tip of the cantilever.
The electromagnet isplaced above the ferromagnetic mass. If AC current is applied to theelectromagnet,
the magnet repeats pulling and releasing the cantilever. If thecantilever is released, the bent cantilever
returns to the initial position by its ownspring force. One implied factor is that if the frequency of current
applied to theelectromagnet equals the resonance frequency of the cantilever, the actuation isamplified by
the Q factor of the mechanical resonance. The resonant frequency frof the cantilever with mass can be
varied due to the change of its mechanicalboundary conditions. The feedback unit consists of three sub-
units, namely thestrain gauge, amplifier and voltage controlled oscillator (VCO). When thecantilever
vibrates at different frequencies, the value at the strain gauge decreases.The drop in the value of the strain
gauge triggers the VCO to adjust the frequencyusing linear feedback to find the new f r(Fig. 2.9).
Fig. 1.1.1. Relation between chip thick-ness and resisting shear stress for carbonsteel SAE 1112.
Fig. 1.1.2. Distribution of defects in materials: failure due to movable dislocations in ductile
materials and fracture dueto microcracks in brittle materials.
For processing units larger than 10 μm, the breakdown of ductile metals due to shear slip begins
at a weak point at a grain boundary or cavity; in brittle ceramics, the breakdown due to brittle
fracture occurs mainly from cracks around the grain boundaries.Therefore machining with the
sharp edge of an ordinary solid tool or the fixed abrasive of a grinding wheel cannot produce
chips of fine atomic-bit sizes, because the cutting edges wear quickly due to high resistive stress,
but diamond tools and abrasives can be used for mirror cutting or grinding because of their
greater wear resistance. Moreover, lapping and polishing using replenishable free abrasives can
be used to realize atomic-bit processing of materials.
Fig. 1.1.3. Elastic failure and fracture limit, : specific stock removal energy (J cm-3) >> s ; :
elastic failure limit at dislocationless region th ; s ps : specific energy for shear slip; tf :
elastic fracture limit without microcrack E / 2 th ; ts : equivalent elastic failure limit in
microcrackless region: shear slip; t pt : specific energy for tensile breakage.
Which corresponds at the microscopial level to the specific volumetric lattice bonding energy
U b (MJ m 3 or J cm 3 ) or atomic bonding energy Eb (J/atom) as given in table 1.1.2. The values
for the abrasives Al 2 O 3 , SiC and diamond are two or three orders of magnitude higher than for
Fe; for atom-cluster processing, the values may be one to two orders of magnitude lower.
(a) Atom-cluster processing with free fine abrasives
Ordinary solid cutting tools and abrasive grinding wheels cannot be used for cutting and grinding
with processing units of atom cluster, because wear of the cutting edge becomes extremely high
for tough stock material such as steel. For ductile light metals such as aluminum, however, it is
possible to realize micro-machining based on atom-cluster processing using diamond tools or
diamond powder (see Fig. 1. 1.4).
Table 1.1.2 Specific volumetric lattice bonding energy U b and atomic bonding energy Eb , and
hardness values of materials*
Material Ub (MJ m-3) Eb (J/atom) (eV/ Remarks Knoop Scratch
atom) indentation hardness Morse
hardness Woodwell
(10 MPa)
Fe 2.6 x 10 3 1.6 x 10 -20 0.1 for tension (200 hardened
700-800)
(1.03 x 103) (8 x 10 -21) 0.05 for shear
SiO2 5 x 102 4.24 x 10-19 2.65 for shear 820
A1 3.34 x 102 2.06 x 10 -21 0.013 for shear
A1 2 O 3 6.2 x 10 5 5.26 x 10-18 32 for tension 1600-2050 9
Si 7.5 x 105 1.59 x 10-17 36 for tension 2400-2550 11
SiC 1.38 x 106 1.1 x 10 -17 67 for tension 2400-2550 14.0
cBN 2.09 x 106 1.07 x 10-17 106 for tension 3000-3200 19.0
B4C 2.26 x 10 6 1.8 x 10 -17 111 for tension 2700-2800 19.7
Diamond I 5.64 x 106 4.5 x 10-17 274 abundant N 8000-8500 71.0
(natural)
Diamond II 1.02 x 10 7 8.2 x 10 -17 513 N-free (5700-10 400) 42.5
In addition to micro-machining, lapping using fine free abrasives of diamond, A1 2O3, SiC, etc.,
and polishing using fine free abrasives of Fe2O3, Cr2O3, CeO2, etc., are widely used in atom-
cluster processing. Lapping abrasives are refreshed and resharpened by crushing during
operation to achieve continuous removal of material, while polishing abrasives burnish under
extremely large shear stresses based on point defects in the atom-cluster range. Thus polishing is
performed with tough and heat-resistant abrasives.
Geometrical surface contours can be precisely shaped by lapping using preformed medium-hard
plate and subsequent polishing using a small soft plate, because the surfaces of these plates wear
little during processing, because of the processing mechanism. Continuous surface processing
with nearnanometre accuracies can therefore be performed.
So far it has been very difficult to generate aspherical and other curved surfaces, because
highlyprecise relative motion between the workpiece and plate has been difficult to achieve.
Recently however, numerical control of fine steps of 0.1 μm has been achieved and as a result,
new fabrication technologies for precision curved surfaces using generating systems with lapping
and polishing tools will soon be developed (see Fig. 1.1.5).
Fig. 1.1.4. Typical processing systems: cutting, (a) Grinding (smoothing), (b) Turning (mirror-
cutting).
(b) Atomic-bit processing using elementary high-energy particles or a concentrated electric
field
In order to perform atomic-bit removal, for which a high-density processing energy of 10 4-
106Jcm-3 is necessary, processing methods using high-energy particles have been developed, in
which a beam of elementary particles such as photons, electrons or ions, chemically and
electrochemically reactive atoms (reactants) or neutral atoms is applied to the processing point
(see Figs 1.1.6 and 1.1.7 and Tables 1.1.3 and 1.1.4).
Processing methods using high-energy particles have pressing resolutions on the atomic or sub-
nanometre scale. However, it is very difficult to position the processing point with nanometre
accuracies, because unlike the situation with solid tools on machine tools, there is no reference
geometrical surface or axis to control the position of the particle beam at such highprecision.
Fig. 1.1.5. Typical processing systems: lapping and polishing, (a) Master plate processing
(sphere and plane), (b) Form-generating.
(a)
(b)
Fig. 1.1.6. Typical processing systems: ion-beam processing, (a) Fine patterning, (b) Aspherical
lens processing.
The reason for this complex behaviour is that materials are not uniform but invariably possess
various defects such as point defects, dislocations, microcracks, boundary cracks, layers
surrounding the crystal grains, etc., which initiate failure or fracture under load. Accordingly,
defects in the work material play a very important role in materials processing such as removal,
deformation and consolidation.
Furthermore, the failure or fracture behaviour of materials is affected by environmental
conditions such as temperature, atmospheric pressure and humidity and the number of repeated
loading cycles (fatigue limit), as will be discussed later.
In subsequent sections, processing mechanisms are discussed according to the type and presence
of imperfections or defects in the materials, in the following order: processing units of atomic
bits in the no-defect region; processing units of atom clusters based on point defects; processing
units of sub-grain size based on dislocations or microcracks; and processing units of multiple
grains based on crystal grain boundaries.
Several examples of photon beam processing using light, laser and radiation beams are shown in
Table 1.4.1 and Fig. 1.4.2. The photon beam projected locally on to the surface of the workpiece
transmits its energy to the outer-shell electron(s) of the atoms. This energy is transformed into
thermal vibration energy or chemical activation energy of the atom that is hit, as shown in Fig.
1.4.3(a) and (b).
However, the thermal energy transferred from a laser photon, as shown in Table 1.4.1, is smaller
than the atomic bonding energy of materials as shown in Table 1.1.2. Accordingly, atomic-bit
processing of an atom by a single photon cannot be performed with ordinary lasers; instead,
photon beam thermal processing is performed, by locally accumulated heat at the processing
point, conducted from the photon beam’s projected area. The saturation temperature Ts (K) due
to concentrated heat at the centre of the projected beam is derived from heat conduction theory as
(see Fig. 1.4.4)
Fig. 1.4.1. Models for electric field evaporation: sharp tip electrode controlled by piezoelectric
stack positioning devicewith sub-nanometre resolution. Pt denotes traction force due to
concentrated electric field on the sharp tip.
Fig. 1.4.2. Laser sources: (a) solid; (b) CO2 gas (coaxial).
Ts Q / ( a ) qa / (1.4.2)
whereQ is the input power (J s-1), is the thermal conductivity (J m-lK-1s-1), a is the radius of
the projected energy beam (m), and q is the input power density, Q / a 2 W m .-2
For
example, the thermal power density q of a photon beam of radius 0.2 μm necessary to achieve the
evaporation temperature of 15000C for stainless steel is ~ 2.5 x 1011W m-2 with = 0.14 x 102 J
m-lK-1s-1. Temperature analysis of the evaporation process due to a projected photon or electron
beam will be discussed in a later section.
Unlike those of lasers, photon beams of X-ray and SOR radiation can achieve direct evaporation
of one atom by one photon, owing to the high energy of the photons, as shown in Table 1.4.1.
Besides thermal evaporation due to photon energy, there are several reactive photochemical
processes such as those used for photosensitive dry plates and photosensitive plastics, in which
reactions are caused by the projected photons, as shown in Fig. 1.4.3(c). The projected photons
penetrate through the surface of a transparent photosensitive material and react directly with
photoreactive atoms in their path. This creates a centre for a latent image, for example an
activated silver atom in silver halide photography or polymerized active atoms in photoresist
plastics.
Of course, the positioning accuracy of the photo-reacted area is limited to the range of the
equivalent wavelength along the photon beam path. Hence, to increase the resolution for
positioning, the argon laser and excimer laser with short wavelengths, e.g. 0.2 μm, are used.
Fig. 1.4.3. Photon energy transfer mechanisms.
Applications for this kind of photolithography are found in the manufacture of IC wafers with
very fine patterns in the sub-micrometre range, and videodisc master plates with very fine
channels and spots of ~ 1 μm. Details are given in Chapter 4. Recently, X-ray and SOR beams
have also been used for photoreactive processes. These beams consists of comparatively high-
energy photons and consequently have very small equivalent wavelengths, of the order of
nanometres. As a result, these beams are used successfully for nano-patterning of LSIs. Using
such radiation beams, the pattern resolution can be improved to the order of 10 nm, but to
achieve high- resolution patterning in the sub-nanometre range, the technology for post-chemical
etching to develop latent images must be improved.
1.4.3 Directional electron beam processing
A focused electron beam processing apparatus is shown schematically in Fig. 1.4.5; the
mechanism of energy transfer from electron to workpiece atom is depicted in Figs. 1.4.6 and
1.4.7. The basic process performed by an energized electron is thermal evaporation of atomic
bits. The accelerated electrons projected on to the workpiece generally transfer their energy to
the outer shell of the atom and increase the thermal vibration energy of the nucleus. Hence
energized electrons can effectively supply the processing energy necessary for thermal
evaporation in a very finely localized portion of the workpiece. However, it is important to
recognize that the projected electron is absorbed mainly in the region of the penetration depth as
shown in Fig. 1.4.7 and not at the workpiece surface. Electron beam processing was initially
developed to form fine patterns on semiconductor wafers and fine holes or textured surfaces on
diamond and other gemstones. This is because the electron has a very small diameter of 2.8 x 10 -
6
nm from the classical viewpoint and a small mass of 9 x 10 -31 kg estimated from classical
theory, so it can be easily focused into a very narrow beam a few micrometres in diameter and
can be highly energized to several hundred kilovolts (1 keV = 1.602 x 10 -16J).
The incident high-energy electron is able to penetrate the workpiece surface through the network
of the atomic lattice structure because its effective diameter is much smaller than the atomic
lattice distance of 0.2-0.4 nm. The incident electrons penetrate through the surface layer to a
depth Rp (see Fig. 1.4.7) where most are absorbed. The experimental and theoretical penetration
depth is
R p 2.2 10 12.V 2 / cm , (1.4.3)
whereV is the acceleration voltage (V) and the mass
Fig. 1.4.4. Input of thermal energy and distribution and rate of change of temperature, (a) Model
for heating: Q, energy input rate; , thermal conductivity; p, mass density; c, specific heat
capacity; t, time; a, radius of energy beam, (b) Temperature distribution and gradient: Ts,
saturation temperature as defined by eqn. (1.4.2); temperature gradient = (0.84 -
0.254) Ts / a 0.6q / . (c) Rate of temperature change at constant input power: tc, characteristic
response time a 2 / k , where Kis the thermal diffusivity / c . (d) Change in temperature
distribution with input power density.
density of the workpiece (g cm-3). For example, the penetration depth in steel for an electron of
energy 50 keV is ~7 μm and in aluminium ~ 10 μm.
As shown in Fig. 1.4.7, the number of electrons absorbed at the workpiece surface is very small;
the depth of maximum absorption rate Xe and the centre of the absorption depth xd are indicated.
Consequently, electron beam processing by thermal evaporation is not suitable for fine
patterning and other fine machining requiring nanometre accuracy on the workpiece surface.
In contrast to the thermal evaporation process, electron beam processing based on reactive
activation by a focused electron beam, as shown in Fig. 1.4.8, is widely used in electron-beam
lithography with nanometre accuracy and resolution. This process is. based on the activation of
electron-sensitive materialssuch as polymers.Atomsof such materialsareactivated by electrons
passing near the nucleus. In other words, the incidentelectronsactivatetheelectron-sensitive atoms
and cause polymerization or depolymerization along the electron path. The process has been
used toachievevery highdimensionalresolutions in the sub-nanometre range, because no thermal
diffusion and little secondary scattering occur. Therefore a finely focused electron beam can
produce patterning with sub-nanometre accuracy using atomic- bit processing based on reactive
activation.
vapour molecules are removed or diffused into the surrounding reacting gas or liquid. Such
chemical -processing of atomic bits occurs uniformly and at random on the workpiece surface to
create a flat, smooth surface.
If the reacted molecules are insoluble or not in vapour form, chemically reactive deposition
occurs on the workpiece surface, but if the reacted or reagent molecules diffuse into the surface
layers of the workpiece and react with the atoms or molecules there, then chemically reactive
surface treatment is performed.The dimensional accuracy obtainable (due to scattering errors) in
chemically reactive processing is in the nanometre range, with stable processing conditions
(temperature and state of turbulence in reagent and liquid flow).To obtain ultra-precision
products with nanometre accuracy (due to deviational errors) by means of chemical processing
or etching, in-process measurement and feedback control of the position of the processing point
and processed volume (area and depth) are necessary. However, this is difficult to realize in
practice.
In general, control of the processing-point position or area in pattemmaking is achieved with a
patterned mask made by the photoresist method. However, control of the processed volume or
depth can be done only by adjusting the processing time and flow rate of etchants. Although this
kind of chemical etching process uses open-loop control without in-process measurement and
control, the dimensional accuracy or precision is fairly high. This is because for chemical
reactions the processing rate is very slow and the processing resolution very fine.To obtain high
processing accuracy, in-process measurement and feedback control with nanometre accuracy
should be developed and applied to chemically reactive processing. Although several in process
measurement and control systems which usesecondary radiation emissions at the point of
chemical reaction are being developed, these are not yet satisfactory.
Fig. 1.4.12. Ion beam source: ion shower type (d.c. discharge and 2.45 GHz microwave
resonator).
Fig. 1.4.13. Molecular beam epitaxy (MBE).
In this section, chemical reactions are treated only at the macroscopic level, dealing with
chemical equilibrium and reaction rates, based on the chemical potential energy due to the
quantum state of electrons in the Bohr model. Microscopic treatment of chemical reactions using
quantum theory, based on the wave function or atomic orbit, is not discussed here, because
atomic-bit processing of materials does not directly concern quantum-mechanical atomic
structure.This section also covers electrochemical processing, because the basic reaction is the
same as in otherordinary chemical processes.
Fig. 1.4.14. Crystal growth by molecular beam controlled by digital computer.
However, the activation energy for the reactions is given by the electric field potential and differs
from the ordinary activation energy in chemical reactions based on thermal potential energy.
Module-II
Besides disturbances, some inherent errors of a machine tool have been shown to be suppressed
in another WORFAC experiment, for cylindrical turning3. The in-process measurement is carried
out at the opposite side from the cutting point, as shown in Fig. 2.1.3. The errors of the slide
motion, such as waviness and lack of straightness, are found to be reduced and converged by
computer simulation as well as a turning experiment using a particular feedback level which is
the detected error multiplied by a certain coefficient (< 1). The corrected contour of cylindricity
is shown to the right of the arrow in Fig. 2.1.4, where the slide motion errors are corrected by the
WORFAC control, while some waviness and inclination can be recognised on the left, the no-
control side of the figure.
Fig. 2.2.1. Zeiss system of stabilization of the interferogram for large optics measurement.
On-machine measurement is intermittent or stop- and; measure, in which the measuring
instrument is placed on or unified with the machine tool. It is important to keep the workpiece
held with a chuck during the measurement so as to be able to reprocess immediately if necessary.
To save time and avoid some errors, the CSSP (Canon super smooth polisher) system has been
developed®, as shown in Fig. 2.2.2.
Fig. 2.2.3. Z-coordinate measurement system of the CSSP.
Fig. 2.4.4. From Talysurf S5 (Rank Taylor Hobson Ltd.) Schematic diagram of laser pickup
Interference profilometers have been gaining wide use in ultra-precision machining shops, the
semiconductor industry and academic laboratories. Although the optical principles and the
mechanical structures differ somewhat, almost all the characteristics and sophisticated computer
algorithms are very similar for both the Mirau(5) and micro-Fizeau(6) interference profilometers.
Typical resolutions in the vertical and horizontal directions are 0.3 nm and 1.5 μm respectively.
A differential measurement method can eliminate common errors and provides a stable signal of
extremely high resolution. Figure 2.5.5 shows a circular scanning heterodyne laser
interferometer^ with better than 0.1 nm vertical resolution. A Wollaston prism divides a Zeeman
split two-wavelength laser beam by polarization into two beams. The measuring beam is focused
on a point on a circular rotating surface, while the reference beam is fixed on the centre of
rotation. The reflected beams are recombined at the prism and create an interference signal
arising from the beam path difference.
The Nomarski-type microscope is recognized as an excellent optical instrument for observing
surface microtexture. An automatic profilometer with the Nomarski prism (8) is shown in Fig.
2.5.6. Two spots focused on the surface are separated by less than the spot size. Since the
interference signal from these spots shows the local inclination of the surface, integration of the
signal produces the surface micro figure. Because of the cancellation of common errors and little
influence of environmental factors, the profilometer achieves a resolution as high as 0.2 nm over
along scanning length, 20 cm.
Fig. 2.5.4. Optical path of the high-precision optical surface sensor (HIPOSS).
Fig. 2.5.5. Schematic diagram of circular-path inter-ferometer.
References
1. Downs, M.J. and Reine, K.W. (1979). A modulated bi-directional fringe-counting
interferometer system for measuring displacement. Precision Engineering, 1, 85-8.
2. Bruning, J.H. (1978). Fringe scanning interferometers. In Optical shop testing, (ed. D.
Malacara), pp. 409-37. Wiley, New York.
3. Koliopoulos, C.L. (1991). Simultaneous phase shift interferometer. Phase Shift
Technology leaflet, 1-9.
4. Kohno, T., Ozawa, T., Miyamoto, N., and Musha, T. (1988). High precision optical
surface sensor. Applied Optics, 27, 103-8.
5. Bhushan, B., Wyant, J.C., and Koliopoulos, C.L. (1985). Measurement of surface
topography of magnetic tapes by Mirau interferometry. Applied Optics, 24, 1489-97.
6. Biegen, J.F. and Smithe, R.A. (1988). High resolution phase measuring laser
interferometric microscope for engineering surface metrology. Proceedings of the
SPIE, 1009, 35-44.
7. Sommargren, G.E. (1981). Optical heterodyne profilometry. Applied Optics, 20, 610-18.
8. Bristow, T.C. (1992). Surface roughness and waviness measurements for optical parts.
Proceedings of the SPIE, 1720, 119-21.
9. Maeda, S., Hiroi, T., Makihira, H. and Kubota, H. (1991). Automated visual inspection of
LSI wafer patterns using a delivative-polarimetry comparison algorithm. Proceedings
of the SPIE, 1567, 100-9.
d Cc Cc
2
(2.6.1)
B V 2
where α is the illumination angle of the probe, V is the full-width half-maximum of the initial
energy of the electrons (V) at the electron source, B is the brightness of the electron gun, which
is proportional to the accelerating voltage V, I is the probe current, is the wavelength of the
electron beam, which is proportional to 1/ V and Cs, Cc are the spherical and chromatic
aberration coefficients respectively. As is clear from the dependence of the above equation on V,
lowering V increases ail the terms except for spherical aberration. With the SEM using a
conventional thermal electron gun (TEG), if the accelerating voltage is lowered to 1 kV, d is ~40
nm. To reduce d at low accelerating voltages, it is necessary to use an electron gun with large B
and small V values and to reduce the Cs and Cc of the objective lens.
(c) Field emission electron gun (FEG)
The field emission electron gun is characterised by high brightness and small energy spread. The
brightness is as high as 108 A cm-2 sr-1 even at 1 kV (in contrast to about ~ 103 for a TEG).
Therefore, under typical operating conditions (I = 10 pA, α = 5 mrad), the first term of eqn
(2.6.1) is 0.4 nm, which can be neglected compared with the other terms. V is ~ 0.3 V for a
1 1
cold FEG and ~0.5 V for a thermal FEG. These values are and of that of a TEG. Hence the
7 4
use of the FEG makes it possible to improve considerably the resolution at low accelerating
voltages.
(d) Improvement of objective lens aberrations
Since the conventional SEM is designed for multipurpose use, its objective lens is not optimal at
low accelerating voltages. If the objective lens is specifically designed for use at low accelerating
voltages, this creates more room for design improvements. For the magnetic lens it is known that
the larger the excitation parameter (J2/V), where J is the number of ampere- turns, the higher is
the lens performance. At low accelerating voltages, a large J2 / V value can be obtained with a
relatively small J. Thus the magnetic circuit can be made thin and small, with a simpler cooling
method or without cooling. Furthermore, an extreme modification of lens shape becomes
possible.
Since the line-width measurement system is mainly intended for measuring the line width
automatically, the wafer is placed horizontally. With this arrangement, it is possible to reduce
drastically the ‘working distance’ (i.e. the distance between the wafer and objective lens), thus
allowing the Cs and Cc values to be reduced to millimetre order. Figure 2.6.1 shows a cross-
section of an objective lens that can be used up to 3 kV. As seen in the figure, the wafer is
immersed in the lens field and the electron beam is focused by the pre-field on to the specimen.
With this lens, J2/V = 269 A2 V-1, Cs = 3.2 mm and Cc = 3.4 mm were obtained. The calculated
probe diameter at 1 kV was 5 nm. In this example, secondary electrons are detected by a detector
placed above the lens. Because of their low energy, the secondary electrons are trapped by the
magnetic force line and effectively transported upward. At the point where the magnetic field
disappears, however, they become dispersed. To prevent this dispersion and guide the secondary
electrons to the detector with no loss, a group of properly designed electrodes is necessary. The
practical use of a high-performance lens such as shown in Fig. 2.6.1 was made possible by the
development of a computer program that permitted optimal design of the transport electrodes®.
Figure 2.6.2(a) shows a high-resolution image of a contact hole obtained with a line-width
measurement system that incorporates the above lens and a thermal FEG. Resist residues on the
bottom surface of the hole are clearly seen. An example of line-width measurement is shown in
Fig. 2.6.2(b).
The morphology inspection system is mainly intended for observing fine structures formed on a
wafer surface. With this system, therefore, it is important that the wafer can be tilted as much as
possible. Although the objective lens of a conventional multipurpose SEM (Fig. 2.6.3) has a
specimen-tilting mechanism, the tilt angle of a large-diameter wafer is limited to ~ 45° at a
working distance of 15 mm. In this case, the Cs and Cc values were 61 mm and 26 mm
respectively. If the working distance is made longer, the tilt angle can be made larger, but this
rapidly increases both Cs and Cc. To solve this dilemma and obtain an objective lens which
allows the specimen to be tilted up to 60° and still have small Cs and Cc values, the ‘conical lens’
proposed by Bassett and Mulvey(4) was put to practical use. Figure 2.6.4 shows a conical lens
with an apical angle of 60°. This lens eliminates the inner polepiece and uses the magnetic field
produced by the coil directly as the lens field. The outer shroud of magnetic material, acting as
the return circuit for the generated magnetic flux, serves to reduce the total ampere-turns and
eliminate the magnetic field leakage out of the lens. As shown in the figure, the profile of the
axial magnetic field distribution has a peak in the vicinity of the coil’s lower end and decreases
gradually towards the upper end. The rate of this decrease can be adjusted by controlling the coil
length along the conic generator and thickness distribution. The wafer is placed slightly below
the virtual apex of the cone, so that it can be tilted up to 60° with a relatively small working
distance. The Cs and Cc values obtained were 33 mm and 15 mm respectively, at a working
distance of 6.5 mm — about half those of a conventional objective lens. The heat generated by
the coil is transmitted to the specimen chamber wall by means of an inner coil former of copper.
The maximum temperature rise of the coil was 30°C at 512 ampere-turns (the focusing ampere-
turns for a 5 kV electron beam).
Fig. 2.6.1. Objective lens for line-width measurement system. Cs = 3.2 mm, Cc = 3.4 mm (at
working distance 2 mm), Jmax = 986 (at 3 kV).
Figure 2.6.5 shows an image of the resist pattern of a 60° tilted wafer obtained with the
morphological inspection system incorporating this lens and a thermal FEG. The side wall of the
resist pattern, which cannot be seen by observation from above, is observed three-dimensionally.
Fig. 2.6.2. (a) High-resolution image of contact hole, (b) Example of line-width measurement,
Fig. 2.6.3. Objective lens of a conventional SEM. Cs = 61 mm, Cc = 26 mm, maximum tilt angle
45° (at working distance 15 mm).
Fig. 2.6.4. Conical lens optimized for use with 60° tilting. Cs = 33 mm, Cc = 15 mm (at working
distance 6.5 mm), Jmax = 512 (at 5 kV).
Fig. 2.6.5. Line and space pattern of resist on 60°-tilted wafer, obtained with a wafer inspection
system incorporating the conical lens shown in Fig. 2.6.4 and thermal FEG (V = 1 kV, working
distance 6.5 mm).
References
1. Ura, K. and Fujioka, H. (1989). Advances in Electronics and Electron Physics, 73, 260.
2. Wells, O. (1974). Scanning Electron Microscopy, p. 75.
3. Munro, E. and Rouse, J. (1989). Journal of Vacuum Science & Technology, B7, 1891.
4. Bassett, R. and Mulvey T. (1972). US Patent No. 3, 707, 628.
5. Szot, J., Hornsey R., Ohnishi, T., and Minagawa, S. (1992). Journal of Vacuum Science
& Technology, B10, 575.
Fig. 2.9.5. Vision images for an ARI robot top, from left camera; bottom, from right camera.
A second example shows how a combination of image processing algorithms is used to detect
small defects on printed-circuit boards. As shown in Fig. 2.9.6, after repeated expansion and
contraction of the image, small defects are detected and eliminated. In this method, the input
pattern creates a ‘pseudo standard’ pattern by itself, which is then compared with the original
input pattern(2).
A third example shows an automatic inspection system using a shadow image method for
inspecting chip components on a printed-circuit board (3). As shown in Fig. 2.9.7, the printed-
circuit board is illuminated alternately from two different oblique directions and the
corresponding pictures are stored in separate memories. Next, the two pictures in the memories
are superimposed and one is subtracted from the other, leaving only the shadow image which
corresponds to the chip mounting error. Using this method, the vast amount of unnecessary
image data of printed patterns and characters on the board is removed, making possible very
rapid inspection as compared with pixel-to-pixel calculation. In this algorithm, grey-level
transformations and pixel subtractions are used effectively.
3.1 Guide systems for moving elements: tool rest and workpiece
For positioning of moving elements at nanometre accuracy, it is necessary to provide guide
systems with sub-nanometre scattering error for the transfer mechanism.
3.1.1 Elastic hinge or spring guide
This system is composed of an elastic element lying between two solid moving elements. The
guiding action is performed by deformation of the elastic element, which is called an elastic
hinge or spring guide. One of the most accurate single-block spring guides for nanometre
accuracy is shown in Fig. 3.1.1, used for the STM (scanning tunnelling microscope).
In this system, the friction accompanying relative transfer motion of the moving elements is
removed, except for the internal friction of the elastic material. Also there is no backlash in the
mechanism joining the elastic element and the two moving elements. As a result, elastic hinge
systems are most favourable for guiding two moving elements at nanometre accuracy. Of course,
wide-range transfer cannot be realized by this system.
Actual examples of spring guides for linear, torsional, and combined motion are shown in Fig.
3.1.2.
and steel guide races. However, in order to realize motion of nanometre accuracy, it is necessary
to maintain the geometrical accuracy of rolling elements: the maximum eccentricities and
diametral accuracies of the balls and rollers must be 10 nm and 50 nm respectively. Moreover,
prestressing should be always applied between the rolling elements and guide races, to remove
backlash between the two components. Of course, in the prestressed condition, the rolling
resistance and also friction increase, but the accuracy of relative linear motion can be maintained
by moving elements of suitable construction for prestressing.
On the microfine motion of rolling contact, experiments reported by the Yoshida
Nanomechanism Organization reveal that, as shown in Fig. 3.1.3, if the rolling displacement is <
100 nm, a linear relation exists between the transfer resistance and the displacement, but in the
case of displacement between 100 and 400 nm, hysteresis occurs, whereas above 500 nm,
ordinary rolling friction takes place.
These phenomena can be understood as follows. On the basis of the Hertz analysis, under perfect
elastic deformation, the rolling resistance becomes zero, apart from the inertial force.
Accordingly, the linear relation at < 100 nm may represent the difference between the resistances
of the loaded and restoring sides of the roller, based on internal friction of elastic materials.
Therefore, even at < 100 nm very slight hysteresis should be expected. At larger displacements,
microslip and also internal friction losses are always to be expected. However, it is confirmed
that in the region < 100 nm, a linear relation is maintained between transfer resistance and
displacement.
Fig. 3.1.4. Guide mechanisms: (a) ball and screw groove; (b) linear V-groove.
The ball screw V groove guide mechanisms shown in Fig. 3.1.4 are of great use, but there is
always a relatively large spinning friction between ball and guide race. Therefore they always
show stick-and-slip phenomena due to driving force and transfer, and also a relatively large
backlash due to erratic geometrical form. As a result, these devices cannot be used for
nanometre-accuracy mechanisms. However, if the geometrical forms of the ball and races are
made to higher accuracy, nanometre-accuracy transfer with these systems will be possible using
lubricants of high oiliness.
Another mechanism, using a knife edge with sharp edges interfaced between two moving
elements, is shown in Fig. 3.1.5. This mechanism, with a hardened knife edge and agate bearing,
has been developed as a very accurate supporting system for microbalances. This kind of knife
edge interface allows very fine linear transfer of moving elements with a smaller range.
Fig. 3.1.5. Knife-edge and bearing guide.
Fig. 3.1.6. Vibration of tables supported on (a) aerostatic and (b) hydrostatic slides.
Fig. 3.1.7. Table-slide guide system, (a) Opposed-pad hydrostatic slide, (b) Arrangement of
hydrostatic pad considering the elastic deformation of the table.
Fig. 3.1.8. Groove compensation hydrostatic slide, (a) Arrangement of grooves, (b) Guideway
feed system, (c) Detail of T-shaped groove.
Fig. 3.1.9. (a) Hydrostatic slide with a diaphragm restrictor, (b) Static characteristic.
Fig. 3.1.10. (a) Hydrostatic slide with a self-controlled restrictor using a floating disk, (b) Static
characteristic.
Fig. 3.1.11. Hydrostatic journal bearings: (a), (b), see text.
(b) Hydrostatic journal bearing
Like hydrostatic slides, hydrostatic journal bearings have various advantages for ultra-precision
machine tools.
Figure 3.1.11 shows typical configurations of hydrostatic journal bearings. The bearing shown in
Fig. 3.1.11(a) is an ordinary bearing which has several pockets and restrictors inside the pockets.
The bearing shown in Fig. 3.1.11(b)(6) has drain holes inside the pockets, and a wider range of
design conditions can be selected by changing the diameter of the drain holes.
(a)
Fig. 3.1.12. Hybrid journal bearings with (a) Rayleigh step, (b) inherent compensation.
However, since in these bearings no hydrodynamic effect exerted by shaft rotation can be
expected in the concentric condition, the stiffness of these bearings depends mainly on the
supply pressure. A higher supply pressure is therefore required to obtain a higher stiffness.
As another means of obtaining a higher stiffness in hydrostatic bearings, the use of the
hydrodynamic effect by shaft rotation is very effective. Figure 3.1.12(a) and (b) shows
hydrostatic journal bearings using the hydrodynamic effect to increase stiffness; these are called
hybrid bearings. The hybrid journal bearing in Fig. 3.1.12(a) (7) has Rayleigh step pads inside the
pockets to use the hydrodynamic effect. The one in Fig. 3.1.12(b)(8) has no restrictors such as
capillaries or orifices but has several shallow pockets whose depths are usually several times the
bearing clearance. In this bearing, stiffness in the non-rotating condition is generated by using
the difference in viscous resistance between the shallow pockets and the lands. Moreover a pair
of these pockets and lands forms a Rayleigh step bearing in the rotating condition and produces
hydrodynamic pressures.
Hybrid bearings can achieve a higher stiffness by using the hydrodynamic effect, but they also
have some disadvantages such as higher power consumption and higher temperature rise under
the rotating condition. To overcome these disadvantages in hybrid bearings, hydrostatic journal
bearings with self- controlled restrictors have therefore been proposed. Since this type of bearing
can achieve a very high stiffness (almost infinite) without using the hydrodynamic effect and a
higher supply pressure, a low power consumption and low temperature rise can be achieved even
in the rotating condition.
Figure 3.1.13(9) shows an example of a hydrostatic journal bearing with a self-controlled
restrictor. The narrow gap hi between the floating bush and the inner bearing sleeve is varied
according to changes in applied load and controls the liquid flow entering the bearing clearance.
This bearing can thus achieve a very high stiffness by adjustment of the pressure P*s. However,
when the applied loads is dynamically changed, this bearing does not always achieve a high
stiffness, and the dynamic characteristics should be taken into account.
Fig. 3.1.15. Groove configurations of aerostatic slides with hybrid restrictors (a, b from ref. 11).
Figure 3.1.16 shows pressure distributions within the slides for different slide clearances. The
variation of pressure distribution in hybrid restrictors is considerably larger than in groove
compensation restrictors. Accordingly, slides with hybrid restrictors can achieve a higher bearing
stiffness. However, since slides with this type of restrictor have unstable regions owing to
pneumatic instability, as shown in Fig. 3.1.17, the design parameters should be carefully
selected.
To obtain a very high bearing stiffness, aerostatic slides with self-controlled restrictors have been
proposed. In fact, some types have almost infinite stiffness. Figure 3.1.18 shows three types of
slide with diaphragm-type self-controlled restrictors. Slides (a) (12) and (b)(13) control the air
pressure distribution within the slide clearance using elastic deformation of
Fig. 3.1.20. Aerostatic journal bearings (b From ref. 17, c From ref. 18, d From ref. 19, f From
refs 20, 21).
their low friction and high rotational accuracy of motion. Figure 3.1.20 shows aerostatic journal
bearings which are often used in actual precision machine tools. Figure 3.1.20(a) shows a journal
bearing with feed-hole type restrictors such as orifices and inherently compensated feed holes.
Figure 3.1.21 compare the dimensionless stiffnesses that can be achieved in these bearings. It
should be noted that the bearing stiffness largely depends on the value of the bearing clearance,
since the dimensionless stiffness includes the clearance.
The rotational accuracy of a shaft supported by aerostatic bearings is greatly affected by the
accuracy of shape of the shaft. Thus a spherical shaft manufactured to a very accurate shape will
have a high rotational accuracy. Figure 3.1.22 shows a spherical-bearing shaft system (22). The
rotational accuracy of this system is < 0.05 μm.
(c) Aerostatic lead screw
Aerostatic lead screws have the same kind of superior characteristics as aerostatic bearings: low
friction, low torque fluctuation and high accuracy of motion. But since they are difficult to
manufacture, they have rarely been applied to precision devices. Recently, aerostatic
in relative motion between magnetic material and a magnetic field. However, in relative motion
of nanometre accuracy, the active system is applied, because the relative transfer range is very
small.
Concrete examples are shown in Fig. 3.1.24 for rotational and linear guide systems. The systems
are based on the magnetic attractive force due to an electro magnetic coil and ferrous material. In
practice, as shown in the figure, two magnetizing coils and positioning sensor coils are provided
to control the gap distance between the electromagnetic coil and moving elements. In these
systems, the attractive forces between two opposed parts are balanced and the relative gaps can
be maintained constant. However, these controlled gaps are always affected by mass and load
variations. Accordingly it appears to be very difficult to maintain nanometre accuracy. Hybrid
systems using ball and roller and magnetic systems as well as pneumatic bearing systems are
used in scanners of A—D converters.
Fig. 3.1.23. Aerostatic lead screw
Fig. 3.1.24. Magnetic guides (active systems): (a) rotary; (b) linear.
References
1. Ikawa, N., and Shimada, S. (1986). Accuracy limitation in ultra-precision metal cutting.
Journal of the Japan Society of Precision Engineering, 52, 2000.
2. Kogure, K., Kaneko, R., and Otani, K. (1982). Characteristics of T-shaped groove
compensation bear-ing, JSME Journal, 48, 583.
3. Osaka, T., Unno, K., Tsubo, A., Maeda, Y., and Takeuchi, K. (1991). Development of
high-precision aspheric grinding/turning machine. In Progress in precision engineering, pp.
Springer-Verlag, Berlin.
4. Ono, K. and Hirose, S. (1985). Static characteristics of self-controlled externally
pressurized thrust bearing with diaphragm restrictor, JSLE, 30, 652.
5. Yoshimoto, S., Anno, Y., and Fujimura, M. (1993). Static characteristics of a rectangular
hydrostatic thrust bearing with a self-controlled restrictor employing a floating disk. Transactions
of the ASME, Journal of Tribology, 115, 307.
Sugita, K. (1984). Ultra-precision bearing. Journal of Japan Society of Precision Engineering, 50,
796.
5. Harada, M., Suda, M., and Miyaji, R. (1985). The improvement of load capacity of
hydrostatic journal bearing using a pad added in the recess (2nd report). Journal of the
Japan Society of Precision Engineering, 51, 841.
6. Qian, D.Z., and Yan, J.F. (1983). ‘DYNASTAT’ bearing-a new hydrodynamic and
hydrostatic hybrid bearing. In 24th MTDR, 361
7. Mizumoto, H., Kubo, M., Yoshimochi, S., Okamura, S., and Matsubara, T. (1987). A
hydrostatically-controlled restrictor for infinite stiffness hydrostatic journal bearing.
Bulletin of the Japan Society of Precision Engineering, 21, 49.
8. Yabe, H., Mori, H., Asakawa, H., and Kihara, M. (1979). Characteristics of surface
restriction compensation aerostatic thrust bearings, JSME Journal, 45, 100-7.
9. Boffey, D.A., Waddell, M., and Deardent, J.K. (1985). A theoretical and experimental
study into the steady-state performance characteristics of industrial air lubricated thrust
bearings. Tribology International, 18, 229-33.
1. 12. Hayashi, K. (1981). Investigation on externally pressurized gas-lubricated circular
thrust bearing with flexible surface. In 8th Gas Bearing Symposium, paper 1. ,
10. Holster, P. and Jacobs, J. (1986). Theoretical analysis and experimental verification on
the static properties of externally pressurized air-bearing pads with load compensation,
Tribology International, 20, 276.
11. Blondeel, E. and Snoeys, R. (1976). Externally- pressurized bearings with variable gap
geometries. In Gas Bearing Symposium, paper E2.
12. Hongo, T., Harada, M., and Miyaji, R. (1987). The static characteristics of externally
pressurized gas thrust bearings with an electrically controlled restrictor. Journal of the
Japan Society of Lubrication Engineers, 32, 894-9.
13. Watanabe, I., Aoyama, N., and Shimokobe, A. (1988). An active air bearing — ultra
precision control of floating position and vibration. Journal of the Japan Society of
Precision Engineering, 54, 329-34.
14. Yabe, H., Mori, H., Aoki, S., and Osame, M. (1985). Characteristics of surface restriction
compensation aerostatic journal bearings. JSME Journal, 51, 3275-80.
15. Ono, K. (1984). Circumferentially grooved hydrostatic gas journal bearing. Bulletin of
the Japan Society of Mechanical Engineers, 27, 95-101.
16. Fujikawa, Y., Yamazaki, S., and Suzuki, M. (1988). Development of high stiffness air
bearing spindle, NTN Technical Review, 54, 65-74.
17. Dee, C.W. and Shires, G.L. (1971). The current state of the art of fluid bearings with
discrete slot inlets. Transactions of the ASME, Series F, 93, 441.
18. Yoshimoto, S., Nakano, Y., and Kakubari, T. (1984). Static characteristics of externally
pressurized gas journal bearings with circular slot restrictors. Tribology International, 17,
199-203.
19. Kobayashi, A., Hoshina, N., Tsukada, T., and Ueda, K., Annals of the CIRP, 27
20. Ishihara, S., Kanai, M., Une, A. and Suzuki, M. (1990). An X-ray stepper for SOR
lithography NTT Review, 2, 92-100.
21. Sakano, T. el al. (1992). Development of nano servo system. FAN AC Technical Review,
5(2), 1-14.
3.2 Servo-control systems for tool-positioning of nanometre
accuracy
3.2.1 Introduction
The concept of position servo-control systems is shown in Fig. 3.2.1. The system consists of a
position command unit, a servo actuating unit, and a material processing unit within nanometre
accuracy, and also in-process position and speed sensing devices with sub nanometre resolution
for feedback control networks with fully closed and semi-closed loops.
In general, tool-positioning on hardware machines tools is carried out by the servo-control
systems based on a software command signal: dynamic motion of hardware of heavy mass and
large braking load is servo-controlled by software information on receipt of a position command
signal. However, position-sensing on 3D measuring machines can now be done in the moving
state using a touch sensor, so the positioning systems are quite different from rigid fixing of the
tool position.
Fig. 3.2.1. Block diagram of servo-control system for tool-positioning. PCD: position command
device PCS: position command signal PC: position compensator kpe: positioning speed signal
SC: positioning speed compensator (0,7: actuator speed signal SD: servo pilot device AD:
actuator for tool-positioning PD: positioning device for tool MD: machining device FW: finished
work SS: speed sensor of actuator, in situ PS: position sensor of tool or workpiece, in situ FS:
form sensor of finished work, post-process AE: actuator driving energy PE: material-processing
energy DT: disturbance SM: stock material PFFS: postion feedforward signal, post-process
PFBS: position feedback signal, in process (semi-closed loop) SFBS: speed feedback signal, in-
process (fully closed loop)
1. The position command unit is composed of a position command device PCD generating a
position command signal, and position compensator generating an actuator positioning speed
signal k p , proportional to the difference between the position command signal and the in-
SD, where is the difference between the command actuator speed signal and the in-situ
actuator speed, performing semi- closed loop control of actuator positioning speed. The servo
pilot device SD provides the drive for the servo- actuator, corresponding to the actuator
positioning speed signal kv .
The elementary servo-control actuating unit may be one of the following types:
(a) mechanical device using friction wheel and control roller systems
(b) hydraulic device
(c) pneumatic device
(d) electric d.c. or a.c. servomotors for stepping and continuous moving systems (rotary and
linear, and direct drive motors)
(e) electromagnetic voice coil device using electric current and magnetic field
(f) electrostatic motor using electrostatic attractive or repulsive force
(g) piezoelectric stack actuator using electric potential control (see Table 3.2.1)
(h) magnetostriction stack actuator using electric current control
(i) ultrasonic driving motor using friction between travelling elliptic wave motion of an
ultrasonically vibrating stack and moving elements
(j) thermal expansion actuator using temperature control
(k) shape memory alloy actuator using temperature control
(l) anisotropic polymer actuator with piezoelectric ceramic stack.
Table 3.2.1 Comparison of actuator characteristicsa
Amplitude Accuracy Working Response
pressure/force time
Pneumatic cylinder 100 mm 100 μm 0.1 N mm-2 10 s
-2
Hydraulic cylinder 1000 mm 10 μm 100 N mm 1s
Voice coil motor 1 mm 0.1 μm 300 N 1 ms
Piezoelectric actuator 0.1 mm 0.1 nm 30 N mm-2 0.1 ms
a
From Piezoelectric actuators, by K. Uchino, Morikita, 1986.
This servo actuating system using a friction wheel mechanism operates as follows. When the
friction roller FR is located at a distance x from the centre of plane wheel PW, the cylindrical
wheel CW is driven through FR by PW. The angular speed of CW is given by
d / dt 0 / R x, (3.2.1)
where 9 is the drive angle of CW, o>o is the constant angular speed of PW, and R is the radius of
CW. Of course, eqn (3.2.1) assumes that the friction forces acting at the contact points of CW,
FR, and PW, under the no-slip condition, are sufficiently greater than the necessary transitional
dynamic forces to accelerate the load L.
The position x of FR is set by the rotation angle of the servo pilot screw SS, as follows:
x P / 2 , (3.2.2)
where p is the length of screw SS. From eqns (3.2.1)
Fig. 3.2.4. Differential planetary bevel gearing. nA : inpilt rotation angle of gear A, n B : output
and (3.2.2) the system gain for the control circuit, given by the ratio of angular speed to rotation
angle of the servo pilot screw, is
d / dt / 0 / R . p / 2 . (3.2.3)
Fig. 3.2.5. Differential planetary plane gearing. A: main coaxial shaft with gear A K: main
coaxial shaft with gear K D: arm shaft with arm G J, B: planetary gears z A , zk , z J , z B : no. of teeth
Therefore if we can set the time constant k or Ts sufficiently small, then the control equation
(3.2.6) tends to the following form in practical use:
(3.2.7)
Accordingly, with this servo-control system, a load of high inertia and large braking torque can
be moved quickly to the command position.
The transient performance of this system is shown graphically in Fig. 3.2.3. From eqn (3.2.6), the
speed of control angle variation is given by
d / dt / K . exp( t / k ) (3.2.8)
Therefore, at the starting point t = 0, the speed of the control angle becomes instantaneously
/ k and decreases to zero. The acceleration of the control angle variation is
d 2 / dt 2 / k 2 .exp t / k (3.2.9)
Therefore, at the starting point, the acceleration becomes instantaneously / k 2 and then tends
to zero.
In practice a step input of angle cannot be performed instantaneously, so the initial speed and
acceleration of the system increase gradually as the increases.
(b) The mechanisms of mechanical compensators using differential bevel and plane planetary
gear are as follows.
The construction of the differential bevel planetary gear is shown in Fig. 3.2.4. The relations
between input and output angles and the differential angle are obtained as follows. When gear B
is fixed,
nB 0, n A / nc 1 i0 , (3.2.10)
and when gear A is fixed,
nA 0, nB / nc i0 1 / i0 , (3.2.11)
where i0 z A / zB . Then the algebraic sum of eqns (3.2.10) and (3.2.11) becomes
n C i0 nB nA / i0 1 (3.2.12)
The construction of the differential plane planetary gear is shown in Fig. 3.2.5. The relation
between input and output angles and the differential angle are obtained by the following
procedures.
With shaft A fixed:
Arm D, nD Shaft K, nK Shaft A, nA
Therefore,
nK / nD 1 i2 i1 / i2 i1 1 k / k , (3.2.14)
Where k i2 i1 , at nA 0
With shaft K fixed:
nA / nD k 1 (3.2.15)
nD nA / k 1 nK k / k 1 , (3.2.16)
c) For mechanical positioning, there are several rigid feedback systems or positively connecting
units using various kinds of mechanical mechanism, for example crank and lever, cam and roller,
and screw and nut, as well as several kinds of gearing, etc. These systems are arranged to
transmit driving power to follower parts and simultaneously to provide positioning information.
Here we describe two kinds of special important positioning systems: indexing mechanisms and
harmonic gearing.
Several kinds of indexing positioning mechanism using a cam and follower are shown in Fig.
3.2.6. These mechanisms are used for intermittent positioning of the worktable. However,
nanometre positioning accuracy cannot be obtained, because neither the mechanical linkage
between cam and follower nor the cam features can be obtained to such accuracy.
A harmonic gearing mechanism to obtain a high reduction ratio is shown in Fig. 3.2.7. A
harmonic gearing consists of three basic elements: wave generator, flex spline, and circular
spline. The wave generator is an elliptical cam with a thin layer of ball bearings on its exterior
and is in general used for input. The flex spline is a very thin deformable cup of high- tensile
steel with fine gear teeth on its exterior and serves in general for output. The circular spline fixed
to the gear case is a rigid hollow ring with fine gear teeth on the inner side, mating with the flex
spline; it has two more teeth than the flex spline has.
The principle of gearing performance is that when the flex spline is deformed elastically by the
elliptical cam of the wave generator, the teeth of the flex spline at the extremities of the cam
mate with those of the circular spline. Accordingly, when the wave generator cam rotates
clockwise by 360°, the flex spline teeth mate in turn with those of the circular spline, and as a net
result the flex spline has rotated anticlockwise by two teeth. Therefore if the total number of
circular spline teeth is 200, then the flex spline rotates by 1/100 per revolution of the wave
generator cam.
This mechanism is very efficacious, because the effective gear reduction is achieved by a plane
gear mechanism with small gear surface slip and as a result, in spite of the very large reduction
ratio, the transmission efficiency is very high. However, although the accuracy of angle
transmission can be expected to be good, at present it is not yet of nanometre-accuracy standard.
where m is the mass of the operating actuator piston rod system with braking load L, and A is the
effective area of the actuator piston. Moreover, the acting fluid flow rate is given by
A dy / dt c pi p1 c p2 pd , (3.2.18)
where c is a coefficient of fluid flow under the linear flow assumption.
Therefore we have the following equations:
p1 pi A dy / dt / c
(3.2.19)
p2 pd A dy / dt / c
pi pd 2 A dy / dt / c m d 2 y / dt 2 L / A, (3.2.20)
If we set the value of total working pressure very high, pi pd m d 2 y / dt 2 L / A , then the
equation reduces to
pi pd 2 A dy / dt / c (3.2.21)
constant of the control system. The system gain dy / dt / 1/ Ts is used to characterize the
system, in general.
Moreover, if for the displacement or position of the actuator rod we form a closed loop by a
feedback lever system,
Y y, (3.2.23)
then the performance of the closed-loop control system becomes
Ts dy / dt y Y (3.2.24)
at the stable condition, dy / dt 0 , and its starting time or response time constant is Ts , which
The positioning accuracy depends mainly on the lap gap width of the servo pilot piston in the
stable state. It is therefore very difficult to make the lap gap width to nanometre accuracy,
especially with a high actuating pressure.
There are also other types of servo pilot valve, such as the D slide and the jet pipe (Askania), and
also several types of actuator such as the great pump, rotary vane, and axial pump types, details
of which may be obtained in the appropriate literature.
(b) A typical hydraulic servo-actuator system applied to an NC (numerical control) machine is
shown in Fig. 3.2.9. It consists of three basic elements: electric stepping motor, hydraulic servo
pilot valve with feedback compensator, and hydraulic actuator of axial piston pump type.
The angular positioning signal delivered from an NC pulse generator is supplied to the stepping
motor SM. Then the analogue output as angular displacement of the stepping motor is transferred
to the piston of the hydraulic servo pilot valve PV through the reduction gear RG. From the other
side, the pilot valve piston is pushed back by the screw thread mechanism of feedback
compensator FC, depending on the angular displacement of output shaft OS.
As shown in section AA of the figure, the output shaft OS is driven by the axial piston pump-
type or swash plate hydraulic actuator SPA, controlled by the servo-pilot valve.
Accordingly, the acting principle of this system seems to be the same as that of the previous
system,
Fig. 3.2.8. Closed-loop control system with rigid feedback network using hydraulic servo-
actuator. PP: servo pilot piston lapping with servo pilot valve port AP: actuator piston with
effective area A (m2) AC: actuator cylinder PR: position commanding rod with displacement
Y(m) AR: positioning rod of actuator with displacement Y(m) FL: position feedback lever of
lever ratio SR: servo pilot valve rod of displacement Y y (m) VP: servo pilot valve
port HS, HR: hydraulic source and drain respectively pi , pd . pressures of source and drain
but the positioning angular resolution attains the order of 10 μm, the limit depending mainly on
the accuracy of the lap gap width.
In addition, the response time constant Ts of the system can be set to the order of 1 ms or the
system gain to several hundred reciprocal seconds. Therefore the driving frequency of the NC
signals can be set to the order of several hundred hertz.
By means of this electro-hydraulic stepping actuator system, the positioning of heavy workpieces
and rigid tools for large cutting forces has been controlled by high-frequency NC signals since
1963. However, recently, this system has been replaced by the d.c. servo motor with high-
resolution encoder, and a.c. invertor servo motor systems are also going to be used.
Fig. 3.2.9. Electro-hydraulic stepping servomotor. SM: electric stepping motor RG: reduction
gear PV: pilot servo-valve FC: feedback compensator with thread mechanism VP: valve plate
AP: axial drive piston OS: output shaft OP: outlet oil port IP: inlet oil port CW: clockwise ACW:
anticlockwise DR: drain SIT: position command signal input terminal SPA: swash plate actuator
Fig. 3.2.10. Pneumatic servo-positioner system with nozzle flap, (a), (b) See text. AS: air source
IO: inlet orifice PR: positioner rod PC: positioner air chamber EM: elastic membrane of
positioner BS: balancing spring of positioner CP: position command pressure pi FN: flap nozzle
FL: feedback lever with flapper PDB: command displacement bellows PDS: command
displacement spring p: positioner pressure equal to flap nozzle p0 . air source pressure : gap
between flap and nozzle Y: position command displacement y: positioner rod displacement L:
load.
3.2.4 Pneumatic servo-actuator systems
(a) The construction of a typical pneumatic servo- positioner system with nozzle flapper is
shown in Fig. 3.2.10a. The system consists of three basic elements: servo-positioner, nozzle flap,
and position command generator.
w i
The position command generator comprises position command signal pressure generator CS,
command position displacement bellows PDB, and spring PDS. The input position quantity Y is
given by the displacement of the end point A of the bellows, according to the pressure pi , of the
input position command.
The nozzle flap mechanism is a highly sensitive pressure transformer, because very fine change
of nozzle gap causes a considerable change in air pressure p in the nozzle. The details of the
mechanism are as follows.
The volumetric flow rate v at the flap nozzle is determined on the assumption that the air flow in
the nozzle is throttled by the area generated by the nozzle periphery d and nozzle gap as
shown in Fig. 3.2.10b:
v c1 d p1/ 2 (3.2.26)
where c1 is a flow coefficient and p is the air pressure in the nozzle. However, v is determined by
the throttling due to inlet orifice IO:
v c2 a p0 p
1/ 2
(3.2.27)
where c2 is a flow coefficient, a is the flow area of the orifice and p0 is the source pressure.
Therefore the relation between the gap of the flap nozzle and the pressure within the nozzle is
Po P / P c1 d / c2 a 2
2
(3.2.28)
dp / d m 2 c1 d / c2 a m Pm2 / P 0
2
(3.2.29)
Then, putting
K m 2 c1 d / c2 a Pm2 / P 0
2
(3.2.30)
dp K m d
around the point m, where km becomes very large at the larger and smaller values of a, which
means that the mechanism can be used as a highly sensitive sensor of fine gap length.
The integrated form in the region of m is
p pm K m m (3.2.31)
and the gap distance of flap nozzle is set by the flap lever as follows:
m Y 1 y (3.2.32)
where is a characteristic constant of the mechanism, determined as shown below, and y is the
output displacement determined by the servo-positioner as
y k p pm (3.2.33)
Fig. 3.2.12. Principle of electro-pneumatic servo-actuator, (a) Self-locking of piston nut. (b)
Positioning of piston nut.
PS: positioning screw rod PN: positioning piston nut SS: screw and nut thread surface section
AC: air cylinder GR: guide rod for piston nut p: air pressure in air chamber : lead angle of
screw thread F: force acting on piston nut N: normal force acting on surface of PS N: frictional
force acting on surface of PS against screw sliding direction P: radial driving force on the surface
of PS due to F R: radius of screw thread pitch : angle of rotation of positioning screw rod PS
NS: assumed surface section of screw thread
shown in the figure, the normal force N acting on the piston nut PN generates a rotation force P
acting along the screw thread surface, given by
P N sin N cos N cos tan (3.2.36)
Accordingly, if tan , then P < 0 (negative). In this case, no rotation of the screw thread PS
can occur i.e. self-locking occurs.
For example, if the friction coefficient μ= 0.2, then the critical self-locking lead angle of the
rectangular thread c = 10°; at = 0.25, c = 14°. For a trapezoidal thread, both and c are
greater than for a rectangular thread.
However, if the positioning screw nut PN rotates positively clockwise by due to the rotation of
stepping motor, as shown in Fig. 3.2.12b, then the piston nut PN moves to the right,
corresponding to the rotation of the screw thread PS. The reason is that the screw surface and nut
surface are separated by the mating clearance between the two surfaces, owing to the force
pushing on the piston nut PN. Therefore a small rotation of the stepping motor is easily
communicated.
By means of this servo-actuator system, a piston nut with a heavy braking load can easily be
moved, aided by pneumatic power, by the command amount of displacement.
3.2.5 Electric servomotor systems
At present the positioning systems of robot hands and tools and workplaces on machine tools are
usually operated by an electric servomotor unit, d.c. or a.c. using a numerical input signal.
(a) Basic electrokinetic motion analysis of d.c. servomotor.
Referring to Fig. 3.2.13a, the basic electromechanical characteristics of a d.c. servomotor are
given by two fundamental equations:
VM N M M (3.2.37)
TM N M iM (3.2.28)
where
VM = induced voltage due to armature winding (V) O>M = angular speed of armature rotor (rad
Accordingly, when the electrical source V0 or input voltage Va is applied to the servomotor with
a rotating load, the following current equations are obtained:
Va V0 R0iM iM Ri VM (3.2.39)
where R0 and Ri are the internal resistances ( ) of the applied electrical source and servomotor
respectively.
Therefore the output torque of the armature shaft is expressed by
TM N M / Ri R0 V0 N M2 / Ri R0 M (3.2.40)
Or
TM N M / Ri Va N M2 / Ri M (3.2.40)
P Ri Re iM2 Ri Re V02 / R0 Ri Re
2
(3.2.41)
Therefore the maximum output power is obtained at matched internal and external resistances of
the cell Ri Rc R0 , because under the resistance- or load matching condition,
dP / d Ri Rc 0 is realized.
Accordingly, when the angular speed of the device is predetermined and the necessary torque is
also limited, a servomotor of suitable electromechanical coupling constant must be selected.
The dynamic kinetic equation of the servomotor unit with load is
J ML dM / dt M TL / n T f / n TM (3.2.42)
where J ML J M J L / n 2 is the total moment of inertia around the armature axis (kg m 2), and
J M is the moment of inertia of the armature itself, J L the moment of inertia of the load itself,
and n the reduction ratio of the gearing, i.e. J L / n 2 is the equivalent moment of inertia of the
rotating load around the armature axis; m is the viscous resistance proportional to the angular
speed of armature and other parts, and T f and TL (N m) are the friction torque and braking
servomotor and tool or worktable for the applied driving voltage is expressed as follows, using
the Laplace notation s d / dt :
(a)
(b)
Fig. 3.2.13. Electric servomotor systems, (a) Circuit network of d.c. servomotor, (b) Transfer
functions. See text for meaning of symbols
As a result, putting T f 0
G p ( s ) M / Va 1/ sJ ML Ri / N M N M (3.2.44)
GL ( s ) M / TL Ri / nN M / sJ ML Ri / N M N M (3.2.45)
Then
Va G p ( s) TL GL (s) M (3.2.46)
d Yd yd (3.2.47)
Then the analogue speed command signal K p is sent to the speed pack SP, which produces the
amplified driving power for the servomotor with voltage Va proportional to the difference
between command speed signal K p and in situ speed feedback signal M . That is,
Va K v K p M K a (3.2.48)
Fig. 3.2.14. Numerical positioning system (NC system) for machine tools PD: position command
pulse rate deliverer Fc: position command pulse rate (10-2mm s-1) PP: position pack FF: feed
forward signal processor Kp: co-efficient of positioning speed proportional to V mm 1 :
difference between command position Y and in situ position y (mm) Yd . digital command position
(mm) yd :digital in situ position (mm) SP: speed pack K v : speed signal amplifying factor K a :
power amplifying factor WU: working unit (servomotor and working device) TM . output torque
of servomotor (Nm) TL : braking torque of load (Nm) Va : driving voltage for servomotor (V) M :
angular speed of servomotor (rad s-1) Gp(s): transfer function for angular speed of servomotor.
Accordingly, if we can set the amplifying ratio term sufficiently large compared with
, K v K a G p S V s rad 1 , the system gain transfer function Gs S tends to the following
Gs ( s) cK p / K s s 1 (3.2.50)
In practice, K s has a nearly constant value of 20-30 s-1 for an SCR (silicon-controlled rectifier)
or thyristor amplifier, and 100-1000 s-1 for a transistor.
Moreover, the system gain K s is inversely proportional to the response time Ts of the worktable
motion. That is, from eqn (3.2.50) with on Gs S dy / dt / , we get the following Laplace
differential equation:
sy K s K s Fc / s y (3.2.51)
where M and L are the angular accelerations of servomotor and rotating load respectively,
With M n L ,
TM n TL / J M n 2 J L / L (3.2.54)
L TM .n / J L J M n 2 (3.2.55)
To obtain the maximum angular acceleration of the load at gear ratio n , make the differential of
L with respect to n zero; then
0 TM J L J M n 2 / J L J M n 2
2
(3.2.56)
Fig. 3.2.15. Block diagram of NC positioning network using servomotor. Symbols as in Fig.
3.2.14.
that is,
J M J L / n2 (3.2.57)
Accordingly, the maximum output angular acceleration is obtained under the condition that the
equivalent moment of inertia of the rotating load is equal to the moment of inertia of the
armature rotor.
(d) Example of a machine-tool position control system
In the systems shown in Figs 3.2.14 and 3.2.15, assume that the feed screw lead l =5 mm and the
gear ratio z2 / z1 n 1 ; then the speed of the worktable is given by
dy / dt cM (3.2.58)
where c = l/2 = 8.0 x 10-4 m. Therefore the equivalent moment of inertia of the worktable
about the armature rotor axis becomes
as deduced from the angular kinetic energy expressed in the following manner:
1/ 2 J LeM2 1/ 2 M cM
2
(3.2.59)
Therefore, as mentioned previously the moment of inertia of the servomotor armature rotor J M
J M J Le (3.2.60)
Hence the servomotor used for this device should have a moment of inertia of the armature
of J M J Le 4.5 10 5 kgm 2 . From the servomotor list of Yasukawa Electric Co., the inertia
motor UGMMEM-06AA having a rotor moment of inertia J M 5.67 105 kgm 2 is selected.
Then the starting time-constant of the motor plus machine tools unit is obtained as follows. The
relation between the generating torque of the servomotor TM and the angular speed M is given
by eqn. (3.2.40');
TM N M / Ri Va N M2 / Ri M
together with the motor data N M 1.06 101 V s rad 1 N m A 1 , Ri 0.84 , and rated input
Combining eqns. (3.2.61) and (3.2.62) and solving with the initial condition M = 0 at t = 0, we
get
Accordingly, the starting time constant Ts of the servomotor plus the machine tool becomes
7.6 103 s at the rated condition, and at the final state the angular speed of the motor becomes
MF 3.8 102 rad s 1 (3.2.64)
The table speed after the starting response time Ts is obtained from the value of Ms
Therefore, in this numerical control system, the minimum time required per unit feed pulse of
0.01 mm must be less than the time constant Ts 7.6 10 3 s , i.e. the minimum pulse width
In the total numerical control system as shown in Fig. 3.2.15, if we assume the peak pulse
voltage to be the rated voltage, 40.5 V, then the system gain K s of the position control system
becomes
K s = (table speed after response time constant/unit pulse feed)
J L mr 2 5 0.3 0.45 kg m 2
2
(3.2.68)
where m is the equivalent mass of the radial arm and grip unit at a mean radius r from the
axis , and m and r are assumed as 5 kg and 0.3 m respectively.
Now if we use a reduction gear ratio n = 100, then the effective moment of inertia J Le of the
radial arm and grip around the servomotor axis becomes
J Le J L / n 2 4.5 10 5 kg m 2 (3.2.69)
Therefore the most effective servomotor is one having an armature rotor with a moment of
inertia nearly equal to the above effective moment of inertia.
In this case, the effective moment of inertia of the rotating load happens to be the same as in
previous example. We can therefore select the same Yasukawa UGEMMEM-06AA motor, and
at the same input rated voltage Va and the same initial condition, the angular speed of the motor
And after the starting response time constant Ts = 7.6 x 10-3 s, the grip unit speed S0 becomes
(a)
(b)
Fig. 3.2.17. Characteristics of servomotor with non-linear driving voltage range, (a), (b) See text
s0 r Ms / n 0.3m 3.8 102 1 0.368 rad s 1 0.72 m s 1 (3.2.71)
Accordingly, for a continuous-path control system, the positioning resolution for a positioning
signal (of pulse width 2Ts becomes
difference falls below the proportional limit 0 , the servomotor is decelerated and reaches the
command stopping position, through the ordinary feedback control network, as shown in Fig.
3.2.17b.
(f) Concluding remarks
There are two other types of servomotors in use: the a.c. rotary or linear stepping motor for open-
loop control, and the d.c. inverter motor for closed-loop control. Details are not given here,
because the operating principle is nearly the same as for the d.c. servomotor dealt with above.
Finally, positioning of nanometre accuracy by a numerical control system using an electric
servomotor has recently been achieved with the Fanuc nano-servo- positioner, as mentioned
afterwards. Also recently, a low-speed and high-torque servomotor, a so-called direct-drive
motor, has been developed. The direct- drive motor will be the most useful servomotor for
positioning of mechanical devices in the near future.
Ms 2
k x Bl i (3.2.72)
The transfer function of the positioner Gex S is given by Gex ( S ) x / E , transformed from
eqn (3.2.74).
Accordingly, for a step input voltage E E 0 , the steady displacement of the moving-coil unit
xs 0 E 0 / Rk / Bl (3.2.75)
That is, the amplification factor K p of the applied voltage for the command displacement X is
K p RK / Bl (3.2.75)
The starting response time for positioning, Tsp, upon a step applied voltage E 0 is obtained from
Lk / Bl Bl s
Rk / Bl x E (3.2.76)
then, referring to the previous section, the response time constant Tsp ( S ) is
for an acoustic transducer, then the response position or amplitude xs for a small angular
frequency becomes
xs E0 sin t / Rk / Bl (3.2.78)
However, when the angular frequency is relatively large, the effect of the natural frequency
n , of the system should be considered. That is, the Laplace- transformed second-order
differential equation deduced from eqn (3.2.74), eliminating the s 3 term, is used as follows:
RM / Bl s kL / Bl Bl s kR / Bl x E
2
(3.2.79)
Then we can transform the above equation to the differential equation as follows:
d 2 x / dt 2 2 dx / dt n2 x E0 / RM / Bl sin t (3.2.79)
Putting x xs sin t and inserting it into eqn. (3.2.79'), we can determine the amplitude
xs E0 / RM / Bl 1 / n2 2 4 2 2
1/ 2
Fig. 3.2.20. Block diagram of complete control system for a positioner. PS position sensor K p =
amplification factor DF: difference finder X: position command signal = correction factor x =
position
Ls R Ms 2 / Bl Bls x E (3.2.80)
MR / Bl s Bl sX E (3.2.80)
Ts MR / Bl
2
(3.2.81)
That is, when MR decreases and Bl increases we can perform quicker positioning in response to a
step input voltage.
The block diagram for the complete control systems for a positioner is shown in Fig. 3.2.20.
3.2.7 Electrostatic servo-positioner
Electrostatic sources utilize the force of attraction between charged capacitor plates. The force is
independent of the sign of the voltage, so a large voltage is necessary for linear operation.
Because the forces are relatively weak, a large area is needed to obtain significant output.
The attractive force can be deduced from the storage energy of a parallel flat-plate capacitor. The
stored energy Es (J) of a capacitor at an applied voltage Va (V), using the SI unit system, is
given by
Es CVa2 / 2
vacuum= 8.854 x 10-12 F m-1, s the relative permittivity of the material, S the area of the flat
plate electrode (m2), and x the distance between the electrodes (m).
Fig. 3.2.21. Model of electrostatic positioner with backup spring, (a) Construction, (b) Balance
of forces (see text). C: capacitor (gas) EL: electrode (applied potential V a) T: moving table
(displacement x) K: backup spring f: attractive force
If we assume a moving flat-plate electrode capacitor with backup spring of spring constant
k N m 1 as shown in Fig. 3.2.21, then the stable balanced position x0 of the moving plate at an
That is,
x0 0 s SVa2 / k
1/3
(3.2.84)
Accordingly, x0 0 s S / k 2 / 3 Va1 Va
1/3
Table 3.2.2 Solid-state positioner or actuator characteristics'1.
Thermal magneto piezoelectric Electrostrictive
expansion striction ceramics ceramics
(metal) (Ni) (general electro (limited electro
strictives, PZT) strictives, PMN)
Hysteresis small small large small
Response s ns-s ms s
time
Driving heat magnetic field electric field electric field
power
Ageing effect small small large small
Strain 105 103 105 103 104 102 109 103
a
From Piezoelectric actuators, by K.Uchino, Morikita, 1986
so x0 / x0 2 / 3 Va / Va (3.2.85)
For example, if we take an air condenser with S 1104 m 2 , Va 500 V, and k 2 103 nm 1 ,
then we get x0 48.0 m and, for V0 10V , x0 0.64 m . Therefore, we can perform fine
position control with an electrostatic positioner, but it is necessary to apply a very high voltage.
Recently an electrostatic servo-motor with specially arranged electrodes and synchronized
exciting voltage has been developed, but it is not yet in practical use.
In this section, however, piezoelectric ceramic servo- positioner units are the main topic. They
are the most widely used in practice, because piezoelectric ceramics can be formed to suit any
required object and can also undergo large deformation with a short response time. Data on
piezoelectric materials are shown in Table 3.2.3.
The basic shapes of piezoelectric ceramic elements — circular plate and cylinder — are shown
in Fig. 3.2.22. Figure 3.2.23 is a graph of the strain- electric field characteristic, showing that the
strain is always positive, regardless of the direction of the electric field, and that there is
considerable hysteresis. Accordingly, when a deformation of positive or negative sign about a
certain point becomes necessary, biased polarization must be effected. Of course, the hysteresis
cannot be removed, but there are several means of overcoming it such as introducing a suitable
capacitor into the circuit.
In addition, there are three types of piezoelectric ceramic element, as shown in Fig. 3.2.24 and
Table 3.2.4. The most important is the piled stack with laminated thin film elements; the single
plate is mainly used in acoustic transducers of Langevin resonance type, and the bimorph
bending plate is used in mechanical switching devices.
The transfer function of an electrostrictive actuator is obtained as follows, referring to the piled
stack form as shown in Fig. 3.2.25. The kinetic equation of the actuator is
M d 2u / dt 2 p du / dt Ac / L u AcdE N (3.2.86)
where
u (t ) = displacement of the centre of equivalent mass (m)
M = equivalent mass of the unit (kg)
p = coefficient of viscous resistance (N s m-1)
A = sectional area of the unit (m2)
c = stiffness of the actuator unit (N m-2)
L = total length of the actuator (m)
d = piezoelectric constant or modulus (strain/ electric field) (m V -1, or C N-1)
E(t) = electric field (V m-1).
Accordingly the transfer function G(s) on u(s) for E(s) is
G ( s ) u / E Acd / Ms 2 ps Ac / L m 2V 1
Therefore, for an applied unit step voltage E s Vm 1 , the displacement u s becomes
u s Acd / M / s p / 2 M 2
2
Table 3.2.3 Characteristics of piezoelectric and electrostrictive materials
Piezo electric Piezo electric crystals
ceramics
General electro Limited electro strictives Unit (dimension)
strictives
Component BaTiO_3 PZT: PMN: ADP Quartz SiO_2
55% PbZrO3 10% PbTiO3 (X-cut)
45% PbTiO3
Compliance s 20 11 13.6 m 2 N 1 1012
(Stiffness) l/S
Relative s 1150 450 15.5 2.5 / 0 1
permittivity
Electromechanic
k31 0.09 [1]
al 0.18 0.20 0.28
coupling factor k33 0.48 0.50 [1]
k
Piezoelectric d 31 60 39 - d11=2.31 mV 1 cN 1 10
constant d 33 140-320 300 dl4 =0.727
1000
modulus)
Mechanical Q 400 [1]
200 103 - 106
quality factor
Curie point Jc 120 290 [°C]
Q = holding energy stored per cycle divided by loss per cycle;
(energy stored mechanically)
k2 d 2 / 0 5 S ;
(total energy stored electrically)
Where 2 Ac / M p / M / 4 s 2
2
(3.2.87)
with the response time constant Ts pL / Ac( s ) , and when the terms in s 2 and s 2 0 , then
u t d / L m 2 v 1 1Vm 1 , m
a
NEPEC - 10, length 20 mm, width 5 mm, thickness 0.5 mm, Va = 100 V
= (590 x 10-12)2 x 5 x 1010/(8.854 x 10-12 x 3400) = 0.58, where k2 is the ratio of mechanical
energy stored in the element to input electrical energy.
Practical examples of piezoelectric ceramic servo- positioners are shown in Fig. 3.2.26. The X—
Y— stage for ultra-precision positioning, Fig. 3.2.26a, is used to obtain a better production
yield in the wafer processing and inspection of semiconductor production, where a positioning
accuracy of 0.1 m and a positioning resolution of 30 nm in the range 30 x 30 m2 is
necessary because the design rule or smallest width of an LSI patterned circuit is 0.3 μm., etc.
The ultrafine controller for mass flow in Fig. 3.2.26b consists of a very fine sensing unit for mass
flow and a very fine valve controller consisting of a piezoelectric ceramics servo-positioner
which is able to control flow rates of up to 20 ml min -1.
Fig. 3.2.24. Different types of piezoelectric (electrostrictive) element, (a) Single plate, (b) Piled
stack, (c) Bimorph (bending) plate, (d) Langevin transducer.
Fig. 3.2.25. Model of piled-stack piezoelectric ceramic actuator, u: displacement of table (m) M :
equivalent mass (kg) Va : applied excitation voltage per pile (V) A: cross-sectional area of stack
(m2) L: total stack length (m) n: number of piles
Fig. 3.2.26. Ultra-precision piezoelectric servo-positioner (from New technology for
piezoelectric ceramics, Ohm-sha, Japan), (a) X-Y- stage for micro-positioning. (b) Mass flow
controller.
3.2.9 Magnetostrictive servo-positioner and acoustic transducer systems
Some ferromagnetic materials become strained when subjected to a magnetic field. The relation
between the static strain and magnetic field is shown in Fig. 3.2.27. As is easily shown, the effect
is quadratic with respect to the field, so a biased field or d.c. bias current is required for linear
operation. Important magnetos-trictive materials and alloys include nickel and permendur (Co-Fe
50 : 50, but fragile). At one time, magnetostrictive transducers for acoustic devices were used
extensively, but they have now been replaced by piezoelectric ceramic transducers.
Magnetostrictive transducers are rugged and reliable but inefficient and configurationally
awkward. Recently it has been discovered that certain rare earth-iron alloys such as terbium-
dysprosium- iron possess extremely large magnetostriction (as much as 100 times that of Ni).
They have relatively low eddy current losses but require a large bias field, are fragile, and have
yet to find significant applica¬tions. Metallic glasses have also recently been considered for
magnetostrictive transducers.
In practice, the static strain due to magnetostriction is top. small for direct application to machine
tool positioning. Therefore an ultrasonic resonant vibrator as shown in Fig. 3.2.28 is used for
positioning, in which the maximum amplitude of the vibrator end can be used as a stopper for
solid tools. The positioning accuracy is inadequate, but the response is quite large.
Numerical data on the characteristics of several magnetostrictive materials are shown in Table
3.2.5.
3.2.10 Ultrasonic servomotor or positioner
Ultrasonic servomotors (Fig. 3.2.29) are realized as stationary-wave and travelling-wave
systems.
The stationary-wave motor system is directly operated by ultrasonic motion of a piezoelectric
oscillator or head of a Langevin-type resonator. A moving or rotating element with lining plate
for friction is in contact with a driving element consisting of an elastic body and a piezoelectric
ceramic plate for ultrasonic vibration. Figure 3.2.30a shows the principle of stationary-wave
generation.
The travelling-wave motion is originated as shown in Fig. 3.2.29b and c, using two layers of
piezoelectric plates. The moving element is driven by the travelling wave through the lining plate
on the elastic body.
Block diagrams for control of an ultrasonic motor or positioner are shown in Fig. 3.2.31. Of
course the positioning accuracy is not very high, because the friction transmission mechanisms
are interlinked.
4 107 H m 1
b
Fracture stress
1/ 2 R L / 2
2
Fig. 3.2.30. Principle of ultrasonic servomotor, (a) Stationary-wave generation, (b) Travelling-
wave source, (c) Travelling-wave generation.
5. Maximum interpretation rate of G-code blocks: assuming a series of 3 μm linear segments and
A maximum contouring control velocity of 180 mm min-1 (= 3 mm s-1), the maximum
interpretation rate of the G-code blocks is
2mms / 3 m 1kHz
1
In other words, one G-code block must be interpreted in <1 ms. This means that NC systems
with
Fig. 3.2.31. Block diagram of ultrasonic servomotor control, (a) For speed control: 1, d.c. source;
2, d.c. voltage control; 3, speed command; 4, speed control circuit; 5, power amplifier; 6,
ultrasonic motor; 7, encoder; 8, feedback circuit, (b) For positioning: 1, d.c. source; 2,
positioning circuit; 3, position command; 4, positioning unit; 5, speed controller; ultrasonic
motor; 7, encoder; 8, feedback circuit; 9, speed signal.
nanometre resolution need a very high data-processing capacity. This is achieved by using
ultrahigh-speed microprocessors or parallel data processors.
(b) Fully closed loop control including machine dynamics
There are two servo control methods for a precise NC system: semi-closed and fully closed loop
controls: Fig. 3.3.3(a) and (b).
In semi-closed loop control, the motor’s position, velocity and current are fed back to the servo-
controller, as shown in Fig. 3.3.3(a). The motor drives the machine through a mechanism such as
a gear and ball screw. Even if the motor’s position can be controlled without, error, errors will
exist in the machine because of friction and lost motion in the drive mechanism. Semi-closed
loop control is therefore not suitable for numerical control systems that need to achieve
nanometre resolution. In semi-closed loop control, however, stability in control and good
operability are relatively easy to achieve.
In fully closed loop control, on the other hand, the machine’s position is directly fed back to the
servo- controller, as shown in Fig. 3.3.3(b). Fully closed loop control therefore provides a better
control performance than semi-closed loop control. It is important to note, however, that fully
closed loop control, which includes machine dynamics such as friction, backlash, and oscillatory
characteristics, is prone to control instabilities. Since these machine dynamics are not negligible
in nanometre-precision machine control, they must be compensated in various ways. Since
backlash cannot be easily compensated, the machine must be designed to avoid backlash by
using such methods as the application of a preload. A preload will increase the friction force. The
friction force can be compensated by a disturbance observer, which is described later in this
section. Oscillatory characteristics can be damped by acceleration feedback. Hence in the design
of a fully closed loop control system, the machine dynamics must be precisely measured and
taken into account.
(c) A.c. servo system by DSP control
There are two types of servomotors: d.c. and a.c. A d.c. servomotor has a commutator and
brushes. Contact between these elements generates a friction force which varies with the motor’s
rotational speed. An a.c. servomotor, on the other hand, has no contacts and no resulting friction.
With similar control performances, therefore, an a.c. servomotor is more suitable to drive a
nanometre-resolution machine than is a d.c. servomotor. A brushless d.c. servomotor can be
considered as a type of synchronous a.c. servomotor.
By using a DSP (digital signal processor), the servo control performance of an a.c. servomotor
can be digitally adjusted to the same level as that of a d.c. servomotor. The DSP is a high-speed
microprocessor specially designed for digital signal processing, such as data processing by
control algorithms.
Figure 3.3.4 shows a block diagram of an a.c. servo control system. The motor is a synchronous,
moving- magnet-type a.c. servomotor. The motor’s rotor has several pairs of permanent magnets.
The stator has three phase windings, U, V, and W, and the current is controlled so that the
current vector is perpendicular to the motor angle, regardless of the rotor’s rotation, enabling the
motor to generate maximum torque. The rotor’s rotational velocity, which is detected by a pulse
generator (PG), is used to control the current phase and motor angle. The servo-controller
generates the torque reference so that the motor angle will match the motor angle reference. The
motor’s electrical angles sin( n ) and sin n (3 / 2) are calculated from the motor angle ,
Figure 3.3.5 shows a servo control system with feedforward compensation and a disturbance
observer. In this system, the motor angle is governed by proportional control and the motor’s
angular velocity by proportional — integral control. The dominant closed-loop dynamics of this
control scheme can be approximated by a second-order lag system with response time r and
damping ratio . With such control alone, response errors due to response delays and load
disturbances are inevitable.
To remove errors due to response delays, therefore, feedforward compensations for acceleration
and velocity are added. The gains of the feedforward compensations are set at r2 for
acceleration compensation and 2 r for velocity compensation. These gain settings compensate
for response delays in the original servo control, making quick and fine control possible. The
response error can be expected to be reduced to less than one-tenth of the original amount by this
method.
Next, systematic errors due to disturbance torque are compensated by adding a disturbance
observer. Using signals on the motor’s angular velocity and motor current, the observer estimates
the load torque. The estimated torque is then subtracted from the motor current reference to
achieve load torque compensation. Compensation by the disturbance observer makes it possible
to achieve fine control while the system remains insensitive to unavoidable load changes present
in precise servo control systems with nanometre resolution.
3.3.2 Software servo systems
Previous servo systems used for industrial robots or precision positioning equipment were
inflexible, as shown in Fig. 3.3.6(1) where neither the position loop
Fig. 3.3.5. Servo control system with feedforward compensation and disturbance observer.
gain or velocity loop gain could be changed after the control system was constructed. This was
because the computer was outside the servo system.
However, to attain faster and more precise positioning, it is necessary to be able to change
parameter gains and/or control modes, because the inertial moment, say, in an industrial robot
changes according to its arm movement. With recent rapid developments in personal computers,
software servo systems such as shown in Fig. 3.3.7(2) have become possible. Here, control modes
as well as some loop gains can easily be changed according to the operational conditions.
In this section we introduce an example of a rapid positioning mechanism with nanometre
accuracy which uses a software servo system developed by the author.
(a) ARV system
In the point-to-point positioning method, a table or stage which must be positioned rapidly is
repeatedly subjected to high acceleration and deceleration; the resulting reaction force of the
inertia of the table causes residual vibrations in the equipment, increasing the settling time.
we devised a mechanism called an anti-residual- vibration (ARV) system, shown in Fig. 3.3.8,
which uses a software servo system to prevent residual vibrations. The table displacement y,
measured by a laser instrument, is fed back to the d.c. servomotor. When the table supported on
the linear ball guideway is accelerated to the right, as in the figure, the sub-table supported on
another linear ball guide begins to move to the left. The sub-table is eventually stopped by the
friction force f h , of the linear ball guideway. The sub table thus absorbs the reaction force of the
table inertia and prevents residual vibrations.
(b) Experimental device
The experimental device is schematically depicted in Fig. 3.3.9(a) and its block diagram in Fig.
3.3.9(b). A command from the computer rotates the d.c. servomotor. When the servomotor
rotates the screw shaft by an angle , the axial force P moves the table by displacement y, which
is then measured by the laser instrument (with a resolution of 2.49 nm). The difference x
between y and the target position x is then registered by the computer and sent to the servomotor.
As reported in our previous paper(2), the reaction force of the table’s inertia causes vibrations in
the base centre and also the floating unit, which is positioned between the table and ball nut to
absorb the radial runout of the screw shaft.
These vibrations result in a long settling time of 1s
Fig. 3.3.6. Conventional digital software servo system.
(i.e. the time required for the table to arrive within 10 nm of the target position). To reduce the
settling time and achieve nanometre positioning accuracy, the ARV system mentioned above is
used. The software servo system is needed since the ARV creates different dynamic
characteristics in the table-guideway-ball- screw apparatus for long and short table travel
distances.
(c) Non-linear elastic characteristic
Without the ARV system, when the motor torque is increased linearly, the table displacement y
also increases, as shown in Fig. 3.3.10(a) and (b). When the table reaches xa = 0.3 or 0.6 μm (at
t t1 or t2 respectively), the motor torque is reduced causing the table displacement y to decrease
after overshooting by a short distance. Fig. 3.3.10(c) shows the relation
between motor torque and table displacement y, derived from (a) and (b) and plotted at 1 ms
intervals. This non-linear elastic characteristic is caused by elastic interactions between the balls
and races in the ball screw and linear ball guide. For xa 0.6 m , the balls seem to be rolling
between points E (where y = 0.4 μm) and F in (b) and (c). Todd and Johnson® have made an in-
depth study of this non-linear elastic characteristic in the 0.4 μm range, and Futami et al. (4) have
reported that nanometre accuracy was obtained by using this characteristic with position
feedback to the linear motor.
(d) Frequency-response test
Figure 3.3.11 shows the results of a frequency- response test using the servo-amplifier voltage Vc
as the input and table displacement y as the output: (a)
shows the result when the output amplitude is > 5 μm (out of the non-linear elastic characteristic
range), while (b) shows the result when the output amplitude is 50 nm (within the non-linear
range). The different characteristics in the two cases explain why the loop gains (proportional
gain k p and integral gain ki ) should be changed when the table displacement moves from the
rough positioning for the long travel range into the short travel range. The linear ball guide for
the sub-table also has a similar non-linear elastic characteristic.
(e) Positioning experiment
1. Without the ARV system. The following control procedure with a software servo system was
used:
i. For rough positioning, the table is moved by PI control.
ii. When the table comes within 50 m of the target position, the proportional gain Kp is
increased to twice its previous value. This brings the table very close to the target position.
iii. In the vicinity of the target position where the non-linear elastic characteristic exists, after
letting k p = 0 the integral gain ki is reduced to half its previous value.
The following results were obtained after 10 mm stroke step responses were repeated 20 times:
mean value of positioning error E 1 second after table start, E = 3.7 nm, and standard deviation
of E, 1.8 nm.
Figure 3.3.12(a) shows the table displacement curve in the vicinity of the target position. We see
that the settling time Ts , for the table to come within ± 10 nm of the target position, is 630 ms. In
Fig. 3.3.13(a), (i) shows the vertical displacement of the cast-iron base centre, yb measured by a
capacitance displacement sensor fixed at the floor (see Fig. 3.3.9(a)), and in Fig. 3 3.13(b), (i)
shows the results of FFT (fast Fourier transform) analysis, where f = 17 Hz represents the
vibration of the cast-iron base, f = 30 Hz is the structure’s resonant frequency, and f = 50 Hz is
due to the twisting vibration of the base.
2. With the ARV system. The table is moved by the same PI control as before (i.e. without
ARV). Figure 3.3.12(b) shows that when approaching the target position, the table displays a
vibrational amplitude of over 0.4 μm between t = 100 ms and t = 300 ms, just as in the absence
of ARV. This vibration causes the sub-table to move in an unstable manner. Moreover,
determination of k p and ki is made difficult because of changes in the friction values f and f h .
To solve this problem, pulse width control (PWC) was used, in which a current pulse (pulse
width t and height h) is supplied to the servomotor. This is shown in Fig. 3.3.14, which also
shows the table displacement yd . Figure 3.3.15 shows the relation between the pulse width t
and table displacement yd after a 15 ms lapse; the dispersion ( : standard deviation) exists
because of the change in friction f.
The software servo system developed by the author has the following features (see Fig. 3.3.16):
i. Rough positioning by PI control (t = 0 to ti ). The table is moved in the long travel range.
ii. Fine positioning by PWC ( t ti to tii ). When the table passes the target position by rough
positioning, the command current i for the servomotor is held at zero for 15 ms. During this time,
the table stops overshooting the target position; the elastic force at the two linear ball guides of
the table and sub-table, causing the non-linear elastic characteristics, is almost non-existent at
equilibrium. After the 15 ms period, the pulse width t is determined from the position
difference x x y as registered in the computer. PWC is repeated until x Ax is < 0.4 μm.
iii. fine positioning by I control t tii . The table accurately reaches the target position by
integral (I)
Fig. 3.3.9. Experimental rapid-positioning device, (a) Schematic view, (b) Block diagram.
Fig. 3.3.10. Non-linear characteristics of ball screw and table guide without ARV system (from
ref. 2).
control ( K p 0 ), which is particularly effective in the narrow range of the non-linear elastic
characteristic.
This software servo system is used for 10 mm stroke step responses, as shown in Fig. 3.3.16: (a)
shows the command current i for the servomotor, (b) the table displacement y, and (c) the
magnified displacement in the vicinity of the target position. After the command of i = 0, PWC is
repeated six times, with t = 5, 5, 3, 1, and 1 ms. After PWC, integral (I) control is used to stop
the table at the target position. In an experiment, the 10 mm stroke step response was repeated 50
times. Figure 3.3.17 shows a histogram of the positioning error E 1 second after table start,
showing a mean of E = 1.0 nm with a standard deviation = 3.2 nm.
(f) Effect of ARV system
Fig. 3.3.12(b) shows the table displacement curve in the vicinity of the target position when the
ARV system is used with a table mass of m = 4.0 kg. The settling time Ts is 285 ms. Fig.
3.3.13(a)(ii) shows the vertical displacement of the base centre Yb , while (b)(ii) shows the FFT
analysis results, where we see that the 17 Hz vibration has been completely eliminated.
Using the software servo system we were able to achieve a mean settling time of Ts = 340 ms (for
50 trials) and obtain positioning accuracies of nanometre order.
References
1. Yaskawa Electric Co. (1992). Introduction to servo techniques for mechatronics. Nikkan
Kogyo Shimbun 97 [in Japanese].
2. Otsuka, J. et al. (1993). Ultraprecision positioning using lead screw drive (2nd report) —
nanometer accuracy positioning. Int. Journal of the Japan Society for Precision
Engineering, 27, 2.
3. Todd, M.J. and Johnson, K.L. (1987). A model for Coulomb torque hysteresis in ball
bearings. International Journal of Mechanical Sciences, 29, 355-65.
4. Futami, S., Furutani, E., and Yoshida, S. (1990). Nanometer positioning and its
microdynamics. Nanotechnology, 1, 31.
5. Yang, S. and Tomizuka, M. (1988). Adaptive width control for precise positioning under
the influence of stiction and Coulomb friction. Transactions of the ASME, Journal of
Dynamic Systems, Measurement and Control, 110, 221.
3.5 Future development of micro-actuators: nano-servo-positioning
3.5.1 Introduction
Microrobots — almost science fiction — do not yet exist but have attracted great interest around
the world, as they are considered to be basic technology for the next generation. It is easy to
understand the advantage of miniaturization of conventional machines. It makes it possible to use
them in confined spaces, and it is expected that they can be mass- produced in a synthesized
manner, integrating microprocessors, sensing devices, and micromechanisms, and controlled as
autonomous agents. These minute autonomous devices, intelligent mechanical systems
combining computer hardware and software, advanced control methods, and microsensors and
actuators cooperating synergistically, will force our conventional way of thinking along new
paths, and in directions that are not totally clear at present. To realize microrobots, there many
unsolved problems remaining, in the areas of micro-actuator and sensing technology,
microfabrication and assembly technol- ogy, energy supply technology, control and
commu¬nication technology, materials technology, and so on (1,2).
Fig. 3.5.5. Response to UV irradiation, (a) Displacement of tip. (b) Surface temperature of test
specimen.
by up to ~ 20 s(16). Figure 3.5.4 shows an experimental device for UV (365 nm wavelength peak
with narrow spectral band width) irradiation of bimorph PLZT (La: PbZrO3:PbTiO 3 3 : 52 : 48).
Radiation was applied for 100 s and then switched off for 100 s. The displacement at the tip of
the PLZT and the temperature of the irradiated surface are shown in Fig. 3.5.5. As an example of
application this actuator, we have developed an optical micro-gripper and an optical mobile robot
with non-contact energy transmission. At present, the response time of PLZT is slow and this
should be improved in future for practical use. We are considering an integrated optical servo
control system with energy and information transmission. PLZT has different optical responses
in terms of the three different effects stated above. Hence, multifunctional use of PLZT is
expected to be developed, not only as an actuator but also as an information transmitter.
(f) Other actuators
Other actuators such as a shape memory alloy (SMA) actuator and a polymer actuator have been
developed, and are promising as micro-actuators.
SMA has a shape memory effect depending on the temperature. On miniaturization, heat
capacity decreases and heat radiation from the surface relatively increases, so an improvement in
the response speed can be expected. SMA has many application examples such as medical
operating tools(17-l9).
Fig. 3.5.6. Active catheter with multiple degrees of freedom for minimally invasive intravascular
neurosurgery.
Figure 3.5.6 shows an active catheter with multiple degrees of freedom using SMA
wires(19). This catheter is made of serially connected units which contain SMA wires in the
lumina, and its multiple degrees of freedom are obtained by series-parallel structure. This
catheter has been developed as a medical tool for the improvement of the operability of
minimally invasive surgery. The SMA actuator is promising for integration with micro-
mechanisms. Thin-film technology for the SMA is under development. The cooling rate is
improved by miniaturization, and a bimetal type which utilizes the difference in thermal
expansion has been proposed.
The polymer actuator has the characteristics of strength against impact, force, and moment, ease
of processing and light weight. As examples of application, a micro-probe using a piezoelectric
polymer actuator, a micro-gripper actuated by a pH-driven film device, and a chemical valve
using a polymer actuator contracted by electricity have been proposed. For the medical drug
delivery, a micro-pump using a thermo- responsive polymer gel and a highly water-absorbent
polymer gel have been proposed(20).
3.5.3 Energy supply method
(a) Classification of energy supply methods
One of the final objectives of microrobotics is to realize an antlike mobile robot which is small
and intelligent to perform given tasks. Most present microrobots are supplied with energy by the
cable. But when a robot becomes small, the cable disturbs its motion with great friction. Flence
the method of energy supply to the micro-actuator becomes important. Methods are classified as
internal supply (internal energy sources) and external supply (external supply of the energy to the
system but without the need for a cable).
(b) Internal supply method
In this case, the energy source is contained inside the moving body. Electrical energy is
frequently used as internal energy, and for this purpose, a battery and a condenser have been
developed. The battery type is good in terms of output and durability, but it is difficult to
miniaturize. Recently a micro lithium battery of micrometre thickness and current density 60 mA
cm-2, and which is rechargeable at 3.6-1.5 V has been developed by thin-film technology(21). As
for the condenser type, an autonomous mobile robot ~ 1 cm 3 in volume (named Monsieur) was
developed in 1992 by Epson, based on conventional watch production technology. It uses a high-
capacity condenser of 6.8 mm diameter, 2.1 mm thickness, and 0.33 F capacity as an energy
source. The electrical capacity of the condenser is small compared with that of the secondary
battery, but this microrobot uses two stepping motors with current control by pulse width
modulation, and can move for ~ 5 min after ~ 3 min charging.
(c) External supply method
In this case, energy is supplied to the moving body from outside.
The following methods can be considered:
(1) optical
(2) electromagnetic
(3) ultrasonic
(4) other.
Method (1) can be classified into (i) optical pressure by irradiating laser beam, (ii) optical energy
to strain conversion using UV irradiation and photostrictive phenomena, and (iii) optical energy
to heat conversion.
Fig. 3.5.7. Optical mobile robot with non-contact energy transmission on an air table.
As an example of (i), remote operation of the micro-object by a single laser beam as a tweezer
has been proposed. As an example of (ii), an optical piezoelectric actuator such as PLZT has
been developed, as already mentioned. As an example of (iii), a low-boiling liquid has been used
with an optical heat conversion material. Utilization of the pyro-electric effect has also been
proposed to supply external energy. Figure 3.5.7 shows a mobile plate on an air table, using
pyroelectric current generated by the temperature difference arising from the heat applied by UV
irradiation(22). Generally, PLZT can be replaced by another pyroelectric element which can
generate current from a temperature change. The field of the mobile plate is made of square
electrodes arranged in a square grid. The electrodes are 1.5 x 1.5 mm wide and are placed at
intervals of 0.5 mm. The field has many holes (diameter 0.18 mm) placed at intervals of 1 mm,
and air is blown through to float the plate. The bottom face of the plate has several electrodes 1 x
1 mm wide placed at intervals of 4 mm. Each electrode is connected to the PLZT. By UV
irradiation, a thrust is generated between the bottom face of the plate and the field, and this can
be used as the driving force for the plate. By the use of the air table, friction is considerably
reduced and even the weak electrostatic force is enough to move it rapidly. In an experiment, it
moved over the field at a speed of 5 cm s-1. Position control of it can be attained by controlling
the irradiation selectively.
As an example of method (2), microwaves, which have been used for non-contact energy
transmission to aircraft and a solar energy generation satellite, have been considered, but there is
few reports on research on their use for micro-actuators. GMA already mentioned can be
considered to be an example of this method.
As an example of method (3), radiation pressure from ultra-sonic waves can be used for the non-
contact operation and driving of an objects. As another example may be considered the
transmission of force through an external medium to a pipeline maintenance pig. In a further
example, selective energy transmission to an elastic object on a vibrating plate has been
proposed.
3.5.4 Conclusion
For nano-servo positioning, synthesis of the actuator, sensing, control, materials, and mechanical
technology is important. This section has focused on and introduced micro-actuators and their
application, together with the energy supply method, which can be the crucial feature for
practical use of mobile
microrobots. To realize a mircorobot, actuator selection and the energy supply method must be
discussed together, depending on the task and purpose. Moreover, the physics dominating a
macro-object is not always the same for the micro/nano-object. The nano-world in particular is
completely different(23). We can observe it by electron microscope or STM/AFM, where
quantum mechanical analysis is required because of the interaction between the molecules and
the electromagnetic wave effect. In the near future, consideration should be given to the method
of control of the actuator depending on the size of the system.
References
1. Fukuda, T. and Arai, F. (1992). Microrobotics - approach to the realization. In Micro
System Technologies 92, pp. 15-24. VDE Verlag.
2. Fukuda, T. and Arai, F. (1993). Microrobotics - on the highway to nanotechnology. IEEE
Industrial Electronics Society Newsletter, Dec., pp. 4-5.
3. Yamada, A. et al. (1983). Optical energy transformation, Society Publication Center (in
Japanese), p. 7.
4. Mehregany, M. et al. (1990). Operation of microfabri- cated harmonic and ordinary side-
drive motors. In Proceedings, IEEE Micro Electro Mechanical Systems.
1. pp. 1-8.
5. Fujita, H. et al. (1988). An integrated micro servosystem. In IEEE International
Workshop on Intelligent Robots & Systems, pp. 15-20.
6. Fukuda, T. and Tanaka, T. (1990). Micro electrostatic actuator with three degrees of
freedom. In Proceedings, IEEE Micro Electro Mechanical Systems, pp. 153-8.
7. Fukuda, T. and Arai, F. (1992). New actuators for high- precision micro systems. H.S.
Tzou and T. Fukuda (eds.), In Precision sensors, actuators and systems, (eds H.S. Tzou
and T. Fukuda), pp. 1-37. Kluwer.
8. Pister, K..S.J. et al. (1990). A planar air levitated electrostatic actuator system. In
Proceedings, IEEE Micro Electro Mechanical Systems, pp. 67-71.
9. Kim, Y.K. et al. (1990). Fabrication and testing of a micro superconductive actuator
using Meissner effect. In Proceedings, IEEE Micro Electro Mechanical Systems, pp. 61-
4.
10. Hatamura Y., and Morishita, H. (1990). Direct coupling system between nanometer
world and human world. In Proceedings, IEEE Micro Electro Mechanical Systems, pp.
203-8.
11. Higuchi, T., Yamagata, Y., Furutani, K., and Kudoh, K. (1990). Precise positioning
mechanism utilizing rapid deformations of piezoelectric elements. In Proceedings, IEEE
Micro Electro Mechanical Systems, pp. 222-6.
12. Fukuda, T., Kawamoto, A., Arai, F. and Matsuura, H. (1994). Mechanism and swimming
experiment of micro mobile robot in water. In Proceedings, IEEE Micro Electro
Mechanical Systems, (to be published).
13. Polla, D.L. (1992). Micromachining of piezoelectric microsensors and microactuators for
robotics applications. In Precision sensors, actuators and systems, (eds H.S. Tzou and
Fukuda, T.), pp. 139-4. Kluwer.
14. Fukuda, T., Hosokai, H., Ohyama, H., Hashimoto. H., and Arai, F. (1991). Giant
magnetostrictive alloy (GMA) applications to micro mobile robot as a micro actuator
without power supply cables. In Proceedings, IEEE Micro Electro Mechanical Systems,
pp. 210-5.
15. Uchino, K. and Aizawa, M. (1985). Photostrictive actuator using PLZT ceramics.
Japanese Journal of Applied Physics, 24, (Suppl. 42-3), 139-42.
16. Fukuda, T., Hatton, S., Arai, F., el al. (1992). Optical servo system using bimorph optical
piezoelectric actuator. In Proceedings, Third International Symposium on Micro Machine
and Human Science (MHS92), pp. 45-50.
17. Ikuta, K. (1988). The application of micro/miniature mechatronics to medical robots. In
Proceedings, IEEE/ IROS, pp. 9-14.
18. Dario, P., Valleggi, R., Pardini, M., and Sabatini, A. (1991). A miniature device for
medical) intracavitary intervention. In Proceedings, IEEE Micro Electro Mechanical
Systems, pp. 171-5.
19. Fukuda, T., Guo, S. el al. (1993). Active catheter system with multi degrees of freedom.
In Proceedings, Fourth International Symposium on Micro Machine and Human Science
(MHS 93), pp. 155-62.
20. Hattori, S., Fukuda, T. et al. (1992). Structure and mechanism of two types of micro-
pump using polymer gel. In Proceedings, IEEE Micro Electro Mechanical Systems, pp.
110-15.
21. Bates, J.B., et al. (1993). Rechargeable solid state lithium microbatteries. In Proceedings,
IEEE Micro Electro Mechanical Systems, pp. 82-6.
22. Ishihara, H. and Fukuda, T. (1993). Micro optical robotic system (MORS). In
Proceedings, Fourth International Symposium on Micro Machine and Human Science
(MHS 93), pp. 105-10.
23. Arai, F., Ando, D. et al. (1995). Micro manipulation based on microphysics. Proceedings
of the International Conference on Intelligent Robots and Systems, Vol. 2, pp. 236-41
Module-IV
In the 1950s, an interferometrically controlled ruling engine was developed at the Massachusetts
Institute of Technology(5). This was designed to measure and control the position of every
groove, with the mercury green spectral wavelength as a standard, and the positioning accuracy
of every groove was improved to nanometre order so that ghosts in the spectral images were
eliminated completely. This interferometric control system was introduced into ruling engines
both in the USA and in other countries, especially using a frequency-stabilized He-Ne laser as a
light source(6-8).
A laser interferometrically controlled ruling engine constructed at the Central Research
Laboratory of Hitachi Ltd is shown in Fig. 4.1.3 and its control system in Fig. 4.1.4. This is a so-
called shaper-type machine tool; while a diamond tool reciprocates to rule the grooves, the blank
carriage is translated continuously by a lead screw for the groove spacing. A laser interferometer
monitors the translation of the blank carriage. A reference signal, which is an ideal fringe signal
when there is no mechanical error in the translation of the blank carriage, is generated
independently and the mechanical translation error is detected as a phase difference between the
reference and the fringe signals. This error signal causes a servomotor to rotate to compensate
the mechanical translation error for the groove spacing. The sensitivity of positioning error
depends on the wavelength of the light source and the detectable phase difference between the
reference and fringe signals. To improve the sensitivity, a multi-reflection laser interferometer as
shown in Fig. 4.1.5 was used for the ruling engine. Here, one fringe signal interval corresponds
to one-eighth of the laser wavelength (~ 0.08 μm) in the carriage translation. As the
Fig. 4.1.6. Interferogram of mechanically ruled diffraction grating (600 mm -1 1st-order): contrast
between purely mechanical and laser interferometric control.
Fig. 4.1.7. Control system of piezoeiectrically controlled ruling engine.
To evaluate the groove shape, especially for the extremely fine and shallow grooves of soft-X-
ray gratings, it is necessary to use either a scanning tunnelling microscope or an atomic force
microscope to measure the cross-sectional profile. A scanning tunnelling micrograph of a soft-X-
ray grating with a groove density of 2400 mm-1 and 1.7° groove angle is shown in Fig. 4.1.10 (14).
The highest groove density obtained so far by mechanical ruling is 10 000 mm -1 (15)
. Here the
grooves are ruled directly on a polished glass surface.
References
1. Harrison, G.R. (1949). The production of diffraction gratings, I. Development of the
ruling art. Journal of the Optical Society of America, 39, 413-26.
2. Hutley, M.C. (1982). Diffraction gratings. Academic Press, London.
3. Strong, J. (1960). The Johns Hopkins University and diffraction gratings. Journal of the
Optical Society of America, 50, 1148-52.
4. Strong, J. (1951). New Johns Hopkins ruling engine. Journal of the Optical Society of
America, 41, 3-15.
5. Harrison, G.R. and Stroke, G.W. (1955). Interferometric control of grating ruling engine
with continuous carriage advance. Journal of the Optical Society of America, 45, 112-21.
6. Harrison, G.R. (1973). The diffraction grating -an opinionated appraisal. Applied Optics,
12, 2039-48.
7. Jarell, R.F. and Stroke, G.W. (1964). Some new advances in grating ruling, replication
and testing. Applied Optics, 3, 1251-62.
8. Loewen, E.G. (1970). Diffraction gratings for
1. spectroscopy. Journal of Physics E: Scientific
2. Instruments, 3, 953-61.
9. Harada, T. and Kita, T. (1980). Mechanically ruled aberration-corrected concave
gratings. Applied Optics, 19, 3987-93.
10. Horsfield, W.R. (1965). Ruling engine with hydraulic drive. Applied Optics, 4, 189-93.
11. Bartlett, R. and Wildy, P.C. (1975). Diffraction grating ruling engine with piezoelectric
drive. Applied Optics, 14,1-3.
12. Takashima, K. and Nawata, S. (1978). Diffraction grating ruling engine with
piezoelectric drive. Japanese Journal of Applied Physics, 17, 1445-6.
13. Harada, T., Taira, H., Kita, T., and Itou, M. (1987). Groove profile measurement of
diffraction grating using electron microscope. SPIE Proceedings, 815, 118-23.
14. Oshio, T., Sakai, Y., and Ehara, S. (1987). Observation of grating surface by the use of
scanning tunnelling microscope. SPIE Proceedings, 815, 124-6.
15. Takashima, K. (1988). Ultra-precision machining of diffraction gratings. Optical and
Electro-optical Engineering Contact, 26, 218-23 (in Japanese).
1 b2 2 b3 3 b4 4
n w w 2 w 3 w ........ (4.1.3.2)
0 R R R
Fig. 4.1.3.3. Numerical control system for variably space grooves.
Fig. 4.1.3.4. Space variation of grating grooves for aberration-corrected Seya- Namioka
monochromator (600 mm-1, R = 500 mm).
Fig. 4.1.3.5. Comparison of spectral images from Seya - Namioka monochromator (600 mm -1, R
= 500 mm) for the Hg lines at 365.0, 365.5 and 366.3 nm.
The ruling parameters for space variation b2, b3, b4,..... and the tool tilt angle for groove
curvature are chosen so as to minimize each aberration term for the concave grating. Here, the
ruling parameter b2 and are related mainly to focal position and amount of astigmatism, b 3, to
coma, and b4 to spherical aberration.
An aberration-corrected concave grating to enhance the spectral image-focusing property of the
Seya - Namioka monochromator has been designed and fabricated (6). This monochromator is an
instrument for obtaining monochromatic light by simple rotation of a concave grating while both
incidence and exit slits are fixed at an angular position of 70° from the grating centre. However,
the monochromator cannot avoid large coma and spherical aberration on the spectral image when
a conventional concave grating with equally spaced grooves is used. The most appropriate
spacing for reducing such aberrations for a concave grating with a radius of curvature of 500 mm
and a groove density of 600 mm-1 at the centre is shown in Fig. 4.3.4. A concave grating with the
ruling parameters to reduce coma and spherical aberration at 400 nm was ruled mechanically.
The spectral images obtained with the aberration-corrected concave grating and a conventional
one are compared using the mercury spectrum; the result is shown in Fig. 4.3.5. The
effectiveness of a space variation as much as 3 nm is clearly seen.
The nanotechnology for obtaining enhanced image-focusing by diffraction gratings using curved
and/or variably spaced grooves has had fruitful results in the field of advanced science and
engineering such as astrophysics(7,8), plasma diagnostics(9,10), applica¬tion of synchrotron
radiation(11,12), and optical communication(13).
References
1. Namioka, T. (1959). Theory of the concave grating. Journal of the Optical Society of
America, $49, 446-60.
2. Flamand, J., Labeyrie, A., and Pieuchard, G. (1969). Diffraction gratings. US Patent
3,628,849.
3. Noda, H., Namioka, T. and Seya, M. (1974). Geometric theory of the grating. Journal of
the Optical Society of America, 64, 1031-6.
4. Harada, T., Moriyama, S., and Kita, T. (1974). Mechanically ruled stigmatic concave
gratings. Japanese Journal of Applied Physics, 14, (Suppl. 14-1), 175—9.
5. Noda, H., Namioka, T., and Seya, M. (1974). Design of holographic concave gratings for
Seya — Namioka monochromators. Journal of the Optical Society of America, 64, 1043-
8.
6. Kita, T. and Harada, T. (1980). Mechanically ruled aberration-corrected concave grating
for high resolution Seya — Namioka monochromator. Journal of the Spectroscopy
Society of Japan, 29, 256-62.
7. Hettrick, M., Bowyer, S., Malina, R.F., Martin, C., and Mrowka, S. (1985). Extreme
Ultraviolet Explorer. Applied Optics, 24, 1737-56.
8. Harada, T., Kita, T., Bowyer, S., and Hurwitz, M. (1991). Design of spherical varied line-
space gratings for a high resolution EUV spectrometer. SPIE Proceedings, 1545, 2-7.
9. Kita, T., Harada, T. Nakano, N., and Kuroda, H. (1983). Mechanically ruled aberration-
corrected concave gratings for a flat-field grazing incidence spectrograph. Applied
Optics, 22. 512-13.
10. Nakano, N., Kuroda, H., Kita, T., and Harada, T. (1984). Development of a flat-field
grazing incidence XUV spectrometer and its application in picosecond XUV
spectroscopy. Applied Optics, 23, 2386-92.
11. Harada, T., Kita, T., Itou, M., and Taira, H. (1986). Mechanically ruled diffraction
gratings for synchrotron radiation. Nuclear Instruments and Methods in Physical
Research, A246, 272-7.
12. Koike, M., Harada, Y., and Noda, H. (1987). New blazed holographic grating fabricated
by using an aspherical recording with an ion-etching method. SPIE Proceedings, 815, 96-
101.
13. Okai, M. and Harada, T. (1991). Novel method to fabricate corrugation for distributed
feedback lasers using a grating photomask. SPIE Proceedings, 1545, 218-23.
4.2 Nano-lithography
4.2. Photolithography
4.2.1 Introduction
Since photolithography was first applied to semiconductor circuit fabrication, the
performance of semi-conductor circuits has been extended up to ULSI. Photolithography is the
most important and key technology in the semiconductor fabrication system.
Photolithographic technology has been improved in accordance with the demands of higher
circuit integration, and now lines of width several hundreds of nanometres have been fabricated
by photolithography. The light source for photolithography mainly determines the resolution
achievable. The wavelength of the light source has been successively shortened, from the g-line
(436 nm) of the high-pressure mercury arc lamp to its i-line (365 nm), the 248 nm radiation of
the KrF excimer laser, and now the 193 nm radiation of the ArF excimer laser.
The optical wafer stepper, referred to simply as the stepper, is used on almost all production lines
for mass production of ULSI as the photolithographic device. The stepper involves many of the
most advanced component technologies, including nanotechnology.
4.2.1.2 Optical configuration of the stepper
The optical system of the stepper is shown in Fig.4.2.1.1. The key component is the
projection lens for imaging the mask pattern on to the wafer with some reduction ratio.
Theoretically, the resolution of the lenses is defined by the formula
R K / NA (4.2.1.1)
where R is the resolution (nm), AT is a process factor, a constant defined by the process of
pattern duplication and equal to 0.8 under production conditions and 0.65 in R & D, is the
wavelength of the light source (nm), and NA is the numerical aperture (= 1/ cos , where is
the angle formed by the optical axis and the outermost light beam to the image). Hence the
resolution of the optical system is higher, the shorter the wavelength or the larger the NA. The
development of lens design and fabrication technology has improved the lens system so that
shorter wavelengths and a larger NA can be used. The performances of typical lens systems
supplied from 1981 to 1993 are shown in Table 4.2.1.1.
Fig. 4.2.1.2. Optical system of water stepper.
A lens system is made up of about 30 component lenses, each of which has a maximum diameter
of 250 mm and a maximum mass of 10 kg. The total mass of the whole lens system is up to 500
kg.
A stepper which uses an excimer laser as the light source for exposure is called an excimer
stepper. The excimer laser is used to obtain higher resolution, having a shorter wavelength in the
far ultraviolet region than the g-line or i-line in the ultraviolet region. The basic configuration of
the excimer stepper is almost the same as the ordinary stepper except for the light source.
A KrF excimer laser ( 248 nm ) is now in use for test production of ICs, and an ArF excimer
laser (193 nm) is under development. It is necessary to involve features such as a narrow
waveband of the radiation generated high stability of light power, and long operating life when
using an excimer laser for photolithography. A kind of optical monochromator is inserted in the
cavity of the laser to achieve a narrow bandwidth. Some characteristics of a typical excimer laser
for photolithography are listed in Table 4.2.1.1. In the application of the excimer laser, optical
spatial pattern noise, called speckle noise, is generated
by small dust particles in the light path, owing to the high coherence of the laser beam (l). A laser
beam direction distributor should be inserted in the light path in the illumination system to
reduce coherence. In some systems, the distributor is a swinging mirror which is driven
synchronously with the excimer laser pulses. Multiple exposure with different directions of
illumination hides the speckle noise pattern.
Shorter wavelengths allow less material to be used in excimer laser projection lenses, owing to
the low transparency of the materials at the excimer laser wavelength. Typical optical glasses-
green soda-lime glass and white crown glass (BK7) - cannot be used in the ultraviolet region.
Quartz (SiO2) and fluorspar (CaF2) are possible candidates.
A combined system using reflecting mirrors has been developed to reduce the number of lenses.
An example is shown in Fig. 4.2.1.2. A problem is the small exposure area with this system, so
scanning methods have to be used, as in subsection 4.2.1.5.
4.2.3 Alignment system
Integrated circuits are fabricated by applying some 10 to 15 different pattern masks for the
multilayered structure. The alignment between a previously exposed pattern on a wafer and the
succeeding pattern on a mask that will be exposed on the wafer is a critical factor determining
the minimum pattern width.
The stepper has a number of alignment systems, each with a certain attainable accuracy. Wafer
pre-alignment is achieved by means of two rolling pins fitting the facet of the wafer within ±3
μm
From phase modulation theory, the phase of the heterodyne frequency f h f h f1 f 2 varies
according to the displacement of the relative positions of the alignment pattern and the laser
beam. On the other hand, the reference signal fr is processed by the reference signal generator as
shown in Fig. 4.2.1.7. The phase difference between fr and fh corresponds to the displacement of
the alignment mark from the fiducial position. The stage is moved to reduce the phase difference
to zero and thus achieve alignment.
This heterodyne method allows greater disturbance to be tolerated from surface roughness of the
mark, low step dimension of the mark, and high reflectivity of materials such as aluminium. The
alignment accuracy has been found to be from 43 to 85 nm for various alignment marks. The
results of an alignment test are shown in Fig. 4.2.1.9 for two types of alignment mark.
DOF k 1/ NA
2
(5.2.2)
Since high-NA lenses are used for high-resolution imaging, focusing is very critical. Line-width
variation with out-of-focus displacement is shown in Fig. 4.2.10. Autofocusing systems are used
in almost steppers. The
Fig. 4.2.8. Principle of LIA optical signal processing.
step-and-scan system is used for the stage system of the stepper. The stage system may be
replaced by a different type, such as an air-bearing stage.
4.2.6 Resolution enhancement technology
To obtain higher resolution with a normal light source and projection lenses, a number of
resolution enhancement methods have been developed.
A modified illumination method has been developed and used in a production stepper(3). When
a narrow line and space pattern is illuminated, transmitted light is diffracted by the slit composed
of lines as shown in Fig.5.2.12(a) according to Fresnel diffraction theory. The diffraction angle
is given by
sin / p (5.2.3)
where is the wavelength of the illuminating light and p is the pitch of the lines and spaces.
If p is larger than the mask-side aperture of the projection lens, zero-order light and diffracted
±1- order light pass through the lens and focus the image of the mask on the wafer. But if p is
smaller than the critical value for the aperture, the ±1-order light beam is stopped by the aperture.
The focused image is impaired because the ±1-order light beam does not contribute to the image.
However, if the illuminating angle is made oblique as shown in Fig. 4.2.12(b), the
of 10 kV, and the data transfer rate 160 MHz. The writing speed of MEBES is four times that of
EBM- 130/40, resulting in a reticle writing time of ~ 5 h for a 64 Mbit DRAM class.
The EBES4 developed by Bell Laboratories is similar in concept, using a thermal field-emission
electron gun(5). The spot size is 0.125 μm, the current density 1600 A cm -2 at an acceleration
voltage of 20 kV, and the data transfer rate 230 MHz.
Another method of achieving a high throughput system is to adopt the variably shaped beam
(VSB) concept. The writing method for the EX-8 (Toshiba Co.) is shown in Fig. 4.3.4, which
adopts VSB, a continuously moving stage, and vector scanning (beam flies from pattern to
pattern(6). The EX-8 has the ability to write a 1 Gbit DRAM-class reticle pattern, because the
address size, corresponding to the beam size in a Gaussian beam system, is 0.01 μm. The EX-8
can generate triangular beams and rectangular beams from 0.1 to 2.56 μm by using a keyhole-
type second shaping aperture. LSI patterns are composed of slanting-angle patterns in addition to
x-y patterns. A conventional VSB-type EB system approximates a slanting angle by small
rectangles. On the other hand, the EX-8 writes a slanting-angle pattern with a combination of
triangular and rectangular beams, which results in a higher throughput than with conventional
VSB-type EB systems. The throughput is ~ 2 h-1 for a 64 Mbit DRAM-class pattern. Moreover,
the EX-8 is equipped with an alignment function using marks on a glass plate, which enables it
to write a Levenson-type phase-shifting mask(7). It is also equipped with a fully automatic glass
plate loading system, as shown in Fig. 4.3.5. A robot takes a glass plate from a cassette magazine
and puts it into the I/O chamber. After the I/O chamber has been evacuated, the glass plate
moves to the loading chamber, from where a shuttle takes it to the writing chamber.
Writing accuracy, as well as throughput, is very important for a reticle writing system. The
required reticle accuracy is indicated in Table 4.3.2. Figure 4.3.6 shows pattern errors appearing
in a pattern written by a VSB and continuous stage-moving system. The patterns dimensional
accuracy is limited by the resist process in addition to problems inherent in EB lithography, for
example the proximity effect and resist heating. Stitching accuracies are composed of shaped
beam stitching, sub-field stitching caused by sub-deflection, and stripe stitching caused mainly
by the main deflection and stage movement. The improvement of stripe stitching accuracy is the
most difficult among stitching accuracies. A stripe stitching error arises from residual distortion
of the main deflector after distortion correction, stage attitude control error, mechanical vibration
between the electron optical column and the substrate, electrical noise, and beam drift.
Moreover, stripe stitching overlaps beam stitching and sub-field stitching. The only way to
reduce stripe stitching error is to discover the origin of the error factors and to eliminate them
one by one. Long-range dimensional accuracy is related to the difference in writing position
from the ideal position. The accuracy has to be better than 5 x 10 -7, since the required long-range
dimensional accuracy is ~ 0.05 μm per 100 mm.
Furthermore, long-term stability of accuracy is required. It needs great effort to maintain an
accuracy of 5 x 10-7 for a long time(8). Many monitoring systems for acceleration power supply,
stage temperature, beam size, beam drift, and so on have been developed. Generally, an EB
system is very complicated, which poses a problem to be solved in the future.
Data conversion from LSI/CAD data to EB system data is essential for large-volume LSI data.
VSB systems are used in parallel to conventional Gaussian beam systems in most mask shops.
Data conversion from conventional Gaussian beam system data to the VSB system data is
therefore necessary. An example of a data conversion system, in which the EX-8 VSB-type
system and the EBM-130/40 Gaussian beam-type system are used, is shown in Fig. 4.3.7.
LSI/CAD data are converted to EBM-format data for the EBM- 130/40. It takes about two days
for 16 Mbit DRAM- class pattern data to be converted from LSI/CAD' to EBM format, because
data compaction using a hierarchical structure of the LSI data cannot be applied effectively to
EBM data. EBM data are converted to VSB data (EX-8 data) in a very short time. The reticle-
making speed is not so high, because it is limited by conversion from LSI/CAD data to EBM
data in spite of the high writing speed of the EX- 8 and high-speed data conversion from EBM
data to VSB data. A high-speed data conversion system from LSI/CAD data to VSB data has
therefore been developed using a hierarchical structure of the LSI pattern data and a parallel
computer processing method(9). The data conversion time is ~ 30 min for a 64 Mbit DRAM-class
pattern.
References
1. Herriot, D.R., Collier, R.J., Alles, D.S., and Stafford, J.W. (1975). IEEE Transactions on
Electron Devices, 3. ED-22, 385.
2. Takjgawa, T., Shimazaki, K., and Kusui, N. (1986). 4. SPIE Proceedings, 632, 175.
3. Owen, G. and Rissman, P. (1983). Journal of Applied Physics, 54, 3573.
4. Gesley, M. (1991). Journal of Vacuum Science and Technology, B9, 1877.
5. Alles, D.S., Biddick, C.J., Bruning, J.H., Clemens, J.T., Collier, R.G., Gere, E.A., et al.
(1987). Journal of Vacuum Science and Technology, B5, 47.
6. Takigawa, T., Ogawa, Y., Yoshikawa, R., Koyama, K., Tamamushi, S., Ikenaga, O., el al.
(1987). Journal of Vacuum Science and Technology, B8, 1877.
7. Levenson, M.D., Viswanathan, N.S., and Simpson, R.A. (1982). IEE Transactions on
Electron Devices, ED-29, 1828.
8. Anze, H., Tamamushi, S., Nishimura, E., Ogawa, Y., and Takigawa, T. (1992). In Digest
of papers, MicroProcess 92, 5th International MicroProcess Conference, p. 132.
9. Magoshi, S., Koyama, K:, Ikenaga, O., Watanabe, S., Saito, T., Ooki, S., and Sakamoto,
S. (1992). In Digest of papers, MicroProcess '92, 5th International MicroProcess
Conference, p. 128.
10. Pfeiffer, H.C. and Langner, G.O. (1978). In Extended abstracts, 8th International
symposium on Electron and Ion Beam Science and Technology, p. 893. Electrochemi¬cal
Society, Princeton, NJ.
11. Yoshikawa, R., Wada, H., Goto, M., Kusakabe, H., Ikenaga, O., Tamamushi, S., et al.
(1987). Journal of Vacuum Science and Technology, B5, 70.
12. Takigawa, T., Wada, H., Ogawa, Y., Yoshikawa, R., Mori, I., and abe, T. (1991). Journal
of Vacuum Science and Technology, B9, 2981.
13. Sohda, Y., Nakayama, Y., Saito, N., Itoh, H., and Todokoro, H. (1991). Journal of
Vacuum Science and Technology, B9, 2940.
14. Berger, S.D., Gibson, J.m., Camarda, R.M., Farrow, R,C., Huggins, H.A., Kraus, J.S.,
and Liddle, J.A. (1991). Journal of Vacuum Science and Technology, B9, 2996.
15. Abe, T., Yamasaki, S., Yoshikawa, R., and Takigawa, T. (1991). Japanese Journal of
Applied Physics, B3, L528.
4.4 Machining of soft metal mirrors with diamond turning
High-precision machining using diamond cutting tools has been used in finishing of soft
metals such as copper and aluminium, for which it is difficult to perform such machining by
grinding, lapping and polishing. This diamond turning technique is finding wide application in
the finishing of parts used in precision machines and optics such as information equipment, laser
machining equipment, and space equipment. Ultra-precision machines have been developed for
mirror-like surface finishing of precision parts and optics by the use of these diamond tools.
Where f is the feed rate (mm rev-1), r is the nose radius (mm), is the nose angle (rad), and is
the working- end cutting-edge angle, the angle between finished surface and end cutting edge
(rad). In the cutting operation, the surface roughness becomes small when the nose radius r is
large and the working-end cutting edge angle is small.
Fig. 4.4.1. Relation between surface roughness and feed.
Figure 4.4.1 shows a comparison between geometrically calculated surface roughness and actual
surface roughness obtained by using a diamond circular tool in machining aluminium at different
feed rates. The depth of cut was chosen as 2 μm, and nose radii for the single-point diamond tool
were chosen as 2 mm and 5 mm. The ultra-precision machine used in these cutting experiments
was designed for practical fly cutting operation. The tool holder can be dynamically balanced,
and the workpiece is mounted in a vacuum
Fig. 4.4.2. Surface roughness on machined surface of OFHC copper (straight diamond tool).
Fig. 4.4.3. SEM photographs of chip formation from OHFC copper machined with a diamond
tool, (a) Continuous chip, (b) Crystal grain on the chip.
chuck on an X-Y table. The spindle is an externally pressurized air bearing spindle constructed
with spherical bearings(1-3). The spindle was rotated at 1500 rev min-1 (cutting speed 350 m min-
1
). Surface
cutting edge angle . Therefore should be as small as possible in order to make the surface
roughness small. Figure 6.1.2 shows examples of surface roughness measurements for OFHC
copper obtained by using a straight-edge diamond tool. Here, was 3.11 rad and was adjusted
to 0.2 mrad; the feed was 0.013 mm rev-1, the cutting speed was 350 m min-1, the depth of cut
was 0.005 mm, and no cutting oil was used. The surface roughness Rmax measured with a
Talystep was 4 nm, indicated by the double circle in Fig. 4.4.1. The actual surface roughness
coincides roughly with the geometrically calculated value. This is so far the best finished surface
roughness achieved with a diamond tool. Smooth surfaces were obtained by using a highly
accurate machine with an ultra-precision spindle and a sharp diamond tool with adjusted to a
very small value.
Fig. 4.4.5. Electron diffraction patterns of machined surface, (a) As machined, (b) After removal
of 0.2 μm of metal, (c) After removal of 0.5 μm of metal.
Figure 6.1.3 shows an SEM photograph of chips produced when machining OFHC copper. This
photograph suggests that the present mirror-like surface machining gives continuous chips by
microscopic cutting instead of burnishing.
Fig. 4.4.6. Polygonal mirror machining arrangement (a) and machined mirror (b).
Generally a damaged layer remains on the machined surface of a metal. Figure 4.4.4 is an optical
photomicrograph which shows a mirror-finished section of OFHC copper machined to 10-20 nm
Rmax surface roughness under the conditions of 0.013 mm rev-1 feed and 5 μm depth of cut. From
this photomicrograph it can be presumed that the crystal structure of the metal surface is very
little disturbed and that only a slight surface damage layer is produced by diamond cutting. This
surface was therefore observed by electron diffraction. Figure 6.1.5(a) is an electron diffraction
pattern of a machined surface. In this pattern, the electron diffraction ring is indefinite; the
crystal structure of the surface appears distorted, showing the presence of strain. Figure 4.4.5(b)
shows the pattern of the machined surface from which 0.2 μm stock was removed by ion etching.
In this figure the electron diffraction ring is clearly seen. Figure 4.4.5(c) shows the machined
surface from which 0.5 μm metal stock was removed in a similar manner. In this case the
electron diffraction ring is much clearer. From this it can be deduced that the mirror-finished
metal surface has a 0.2 μm surface damage layer.
Fig. 4.4.7. Interference pattern of machined surface edge obtained with an interference
microscope (wavelength 660 nm).
4.5 Mirror grinding of ceramics
Grinding, as well as cutting, is a typical machining method to meet the requirements of surface
accuracy and quality. If the depth of tool action against the work surface is gradually brought
closer to the limit in such machining, and it finally becomes possible to achieve removal of
nanometre-size material, approaching the depth of a single-atom layer, surface accuracy and
quality will be further enhanced. Also, when such machining conditions are applied to hard and
brittle materials, a significantly mirror-like surface can be obtained, without the microcracks
produced in conventional polishing.
Let us assume that constituent atoms are individually removed as the smallest chips from the
work surface. Whether chip formation is effected mechanically, physically or chemically in such
a case, the energy consumed should not differ much if the work consists of the same material.
Therefore it may be said that there is no difference in the material removal mechanism between
the grinding and cutting methods in nanometre-order chip formation. In practical nano-cutting, a
single point diamond tool has been used, while an abrasive wheel on which there is an irregularly
shaped edge of randomly distributed grains has to be applied in nano-grinding. Although there
seems to be a great difference between the two methods, studies on micro- or nano-behaviour in
the cutting edge region of single-point tool cutting have provided strong evidence as to the
mechanism of mirror grinding.
4.5.1 Nano-grinding requirements
To ensure a mirror-like surface on certain kinds of materials, polishing was the primary
machining method at one time. The mirror cutting method for soft and non-ferrous metals, such
as aluminium alloy or oxygen-free copper, was developed in response to the requirement for
fabricating devices in the electronics and optics fields. The material removal mechanism in nano-
cutting to obtain mirror-like surfaces was clarified, and this method has subsequently played an
important role in the fabrication of various high-tech devices.
Nano-cutting can be used for mirror finishing not only of such soft metals but also of polymers
such as PMMA, amorphous metals such as electroless Ni- plated surfaces, single crystals and
polycrystals of KDP, Ge, and Si, or amorphous materials including optical glass. Nano-cutting
methods have made such remarkable progress, that the surface roughness R max of hard and brittle
materials has been reduced to several nanometres. The realization of such nano-cutting has been
supported by the use of ultraprecise and highly rigid cutting equipment, sharpened single point
diamond tools, and a variety of peripheral technologies. This implies some points of similarity in
realizing nano-grinding.
The surface roughness of worked faces is conceptually formed as the profile resulting from the
production of chips, and a sharpened cutting edge is necessary for obtaining mirror surfaces.
Diamond single crystal as a tool material is very hard, the cutting face and flank are polished
smoothly, and the cutting edge formed by their intersection can be extremely sharp.
In the indentation hardness test on a specular glass surface, when an indenter with minimal tip
radius is gradually forced into the surface, the behaviour of a minute area on the glass comprises
elastic deformation at first, then plastic deformation, and finally micro cracking, as is well
known(1). Accordingly, when a hard and brittle material is to be cut to a mirror-like surface as in
cutting aluminium alloy, the desirable working condition is plastic behaviour over a minute area,
for instance 1/10-1/100 of the actual depth of cut for metals, to avoid crack generation (2).
Figure 4.5.1 shows a model of the turning or fly cutting of a hard and brittle material (3). In the
area where the actual depth of cut becomes considerable, the removal of material in the brittle
mode is unavoidably accompanied by some cracking. In the limited area of very small cutting
depth, however, the plastic behaviour in the ductile mode occurs. It is important in nano-cutting
or the nano-grinding to avoid cracks completely, i.e. to adopt a working condition that eliminates
the brittle mode.
When a tool is used with a negative rake angle, e.g. -10° to -30°, for a hard and brittle material,
notable cracks are found to develop in front of the tool edge. Taking account of these results in
nano-cutting, nano-grinding has developed as a more practical technology.
Fig. 4.5.1. Machining geometry for cutting model: cross-section along the cutting direction, with
effective cutting depth increasing with distance from centre of tool. Ductile response is obtained
when the depth of damage, d c , initiated at cutting depth yc does not extend below the plane of
the cut surface.
The need for nano-grinding becomes particularly pressing in fabrication of devices from a hard
and brittle material. In cutting a hard material, wear and chipping of the artificially sharpened
cutting edge are unavoidable problems in practice. When nano-cutting is used for the fabrication
of a large area of several optical devices, a fatal problem such as having to change the cutting
tool halfway through working can occur. When applying nano-cutting to the fabrication of hard
and brittle material devices, the continuous use of single point tool has been found to be difficult.
A grinding wheel, on the other hand, has innumerable tips of fixed abrasives and can maintain an
equilibrium state of budding, flattening and dislodgement of the grains for a fairly long time.
Since it will be possible to obtain a satisfactory mirror-like surface under such a condition, nano-
grinding can be expected to be a valuable mirror finishing method for hard and brittle materials.
In cutting or grinding systems, various elements such as air bearings, hydrostatic pressure
bearings, piezoelectric actuators, and friction drives have been investigated by various
organizations to achieve higher precision and rigidity of movable units. Concerning the materials
of construction of the machines, consideration has been given to thermal expansion and external
or internal vibration; super-invar, low- thermal-expansion glass, and granite have therefore been
used. Furthermore, various improvements including a new cooling system, and ultra-precision
laser measurement system, and a high speed response control system have been adopted to
upgrade grinding technology.
In order to develop the next generation of ultra precision machines, cutting and grinding
systems have been thoroughly reviewed, and several new machine forms have been proposed.
For instance, the tetraform structure, which is in the form of a tetrahedron and consists of six
supported arms, has been reported as well-balanced and subject to minimal distortion under
forces and heat generated during machining, taking account of the fact that there can be no
completely rigid structural material for machines (4,5). In the ductile mode of grinding, a quartz
crystal has been finished to a mirror-like surface with a surface roughness R a of 2.76 nm. In
grinding instead of conventional lapping of silicon wafers, ultra-precision grinding equipment
with a loop stiffness between grinding wheel and work table of 150 N/μm was developed and a
flatness of TTV < 0.6 μm was obtained on 150 mm wafers (6).
In general, a metal-bonded wheel with small diamond grains, 5 μm in size is not practical
because mechanical dressing is usually applied without sufficient consideration and leads to the
dislodgement of the diamond grains. In the Elid grinding method, however, effective dressing is
possible even if the diamond grains on the wheel surface are 1 μm in size. During electrolysis,
the bond material on the wheel surface does not simply dissolve into the grinding solution but is
converted to oxidized or hydroxide films. Therefore, in the application of the Elid method to
nano-grinding, metal-bonded wheels consisting of rather smaller grains are used. The work-
material-removing action, however, should be interpreted to be done by the non-metal bonded
wheel.
The Elid grinding method can be carried out on rotary grinding and reciprocal grinding machines
of the vertical of horizontal type and can be applied to nano-grinding of crystalline materials,
such as silicon wafers and GaAs wafers, in addition to various kinds of advanced ceramic
elements and glass lenses, and enables these materials to be finished to mirror quality with a
surface roughness Rmax of several nanometres.
A comparison between conventional mechanical dressing and the Elid method has been carried
out in the grinding of BK7 glass. As shown in Figure 4.5.3,
when using a mechanically dressed grinding wheel, in spite of a rather smaller grinding force,
the ground surface is covered with microcracks and has a surface roughness of Ra 460 nm. On
the other hand, in the Elid grinding method the grinding force is ten times as high, a surface
roughness Ra 11 nm is obtained, and microcracks are never observed(10). Although the Elid
grinding method should satisfy the excellent nano-grinding conditions, it is important to watch
out for the tendency of the grinding force to rise and to take appropriate action.
In nano-grinding, some grinding wheels made of cerium oxide powders or silicon oxide powders
have been successful in special cases, in addition to diamond grinding wheels. As grinding
wheels of fine abrasive powders can give smaller surface roughness, a method of manufacturing
excellent grinding wheels by the dispersion of abrasive powders has been devised, utilizing
electrophoretic deposition principles. Colloidal silica (10-20 nm) is mixed with sodium alginate
solution, and the abrasive powder is collected at the positive electrode by means of
electrophoretic deposition, then dried, and pelletized into grind stones. By application of a wheel
of such stones to silicon wafer grinding, a mirror-like surface with a surface roughness R max 10
nm and without grain tracks is obtained with 2 μm depth of cut and 800 m min -1 wheel speed(11).
The use of poly (vinyl alcohol) as a bond material instead of sodium alginate for fine diamond
powders has been proposed.
Fig. 4.5.3. Grinding characteristics in (a) mechanical and (b) electrolytically assisted diamond
grinding of BK7. Figures in circles are surface finish Ra (nm); figures in parentheses are normal
grinding force (N) during final cuts.
References
1. Taniguchi, N. Chapter 1 of this book.
2. Miyashita, M. (1988). Review of ultraprecision grinding technology. In Proceedings of
the International Congress for Ultraprecision Technology, Aachen, 41-57.
3. Blackley, W.S. and Scattergood, R.O. (1991). Ductile- regime machining model for
diamond turning of brittle materials. ,13, 95-103.
4. Lindsey, K., Smith, S.T., and Robbi, C.J. (1988). Sub-nanometre surface texture and
profile measurement with NANOSURF 2. Annals of the CIRP, 37, 519-22.
5. Anon. (1991). Design news: Tetraform system revolu-tionizes machine design. Precision
Engineering, 13, (1), 95-61.
6. Abe, K.., Yasunaga, N., Miyashita, M„ Yoshioka, J., and Daito, D. (1993). Development
of ultra precision grinding equipment for ductile mode surface finishing of brittle
materials. In Proceedings of the 7th International Precision Engineering Seminar, 153-64.
7. Suzuki, K., Uematsu, T., Yanase, T., Honma, M., and Asano, A. (1991). Development of
a simplified electro¬chemical dressing method with twin electrodes. Annals of the CIRP,
40, 363-6.
8. Ball, J.M., Murphy, N.A., and Shore, P. (1992). ‘Ductile’ mode diamond grinding of
optical glasses using electrolytic techniques. In XVI International Congress of Glass, Vol.
6, 259-64.
9. Ikeno, J. and Tani, Y. (1990). Nanometer grinding using ultrafine abrasive pellets —
Manufacture of pellets applying electrophoretic deposition. Annals of the CIRP, 39, 329-
32.
4.6 Ultra-precision block gauges
4.6.1 Introduction
Block gauges have several features: they possess a high accuracy as end standards; the
dimensions remain stable over long periods; any arbitrary gauge dimension can be built up by
wringing together several gauges; various applications are possible by combining them with
auxiliary parts; they are easy to use; etc. Currently they are widely used in the machine industry
as length standards.
In manufacturing plants, various length-measuring instruments are used to inspect the
dimensions of fabricated parts and products. To guarantee the accuracy of such length-measuring
instruments, they must periodically be compared with block gauges and calibrated. In turn, the
block gauges used for inspection must be checked against a set of higher- grade calibration block
gauges, and the highest (00)- grade block gauges must be calibrated by absolute measurement
with an optical interferometer. The optical interferometer is a length standard that uses the
wavelength of a standard light (see Section 6.4.3 below). The measured results using the
standard light are traceable to national as well as international length standards.
Fig. 4.6.1. Processing mechanism for chemical polishing of CdTe single crystal, (a) Effect of DN
agent, (b) Effect of oxidizing bleach, (c) Effect of mechanical work by abrasive.
To meet such specifications and applications, the dimensions, flatness, parallelism and surface
rough- ness of the end surfaces of block gauges must be finished to the highest accuracy. Since
their invention, improvements have been made in the material characteristics, heat treatment
methods, and methods of measurement. We briefly discuss below the accuracy, methods of
measurement and fabrication block gauges.
4.6.2 Accuracy of block gauges
The length of a block gauge is defined in ISO 3650 (3) as follows:
The length of a block gauge at a particular point of the measuring face is the perpendicular
distance between this point and a rigid plane surface of the same material and surface texture
upon which the other measuring face has been wrung.
Dimensional measurements of block gauges are thus based on this definition. Since the
advantage of block gauges lies in being able to create any arbitrary dimension by wringing
together individual blocks, it is important to improve the ‘wrung’ strength and minimize
dimensional errors due to the combination.
Flatness, parallelism, and surface roughness are thus related to the accuracy of combined block
gauges as well as to the final dimensional accuracy. The dimensional tolerances and permissible
variations in parallelism are shown in Table 4.6.1 and Fig. 4.6.2 (Detailed specifications are
omitted here.)
4.6.3 Measurement of block gauges
If the block gauge is to be used in industry as a length standard, it must be measured according to
the definition of a metre given by national and international standards. In 1889, the metre was
defined as the length of a prototype standard; since 1983, it has been established as the distance
that light travels in vacuum during 1/299 792 458 of a second (4). The dimension of a block
gauge, on the other hand, is defined in ISO 3650 by two methods: by interferometry and by
comparison with a reference block gauge, as shown respectively in Figs 6.4.3 and 6.4.4. The
former method is used to measure the dimensions of highly accurate grade-00 gauges and
96 198 114
reference block gauges, Kr, Hg, and Cd, recommended by the 17th CGPM (1983) as
standard light sources to realize the definition of the metre, are also used as light sources process
of for block-gauge interferometry.
4.6.4 Measurement by interferometry
There are two methods of measurement using interferometry: the coincidence method and the
counting method. The former was developed early on and is still in wide use today. Although we
do not discuss its principles here, Fig. 4.6.5(5) shows the construction of a block gauge
interferometer developed jointly by the National Research Laboratory of Metrology of Japan and
Tsugami Corp.
In the coincidence method, the fraction of the interference fringes is measured to determine the
block gauge’s dimensional error. The measured result is then corrected by incorporating the
following factors to obtain the ‘true’ dimensional error:
(1) temperature measurement to correct for thermal expansion (thermometer resolution
0.01 K when the coefficient of thermal expansion of the gauge material is 11.5 ± 1.0 x
10-6 K-1);
(2) measurement of characteristic indices of ambient air to correct for variations in the
index of refraction;
(3) correction to account for the size of the pinhole or slit of the interferometer collimator;
(4) correction to account for differences in the base plate material (e.g. glass, fused
quartz).
4.6.5 Fabrication
Although the block gauge is very simple in structure, several processing steps are needed to
satisfy such requirements as dimensional accuracy and long-term stability(6). In addition, to the
traditional materials
4.6.6 Lapping
Lapping is a critical process that determines the dimensional accuracy of the block gauges. This
entire process is carried out in constant-temperature rooms (20 ± 0.2°C), with separate rooms
reserved for rough and finish lapping.
In the early days, the accuracy of block gauges was achieved by hand lapping. A set of eight
gauges was
used, four attached on each side of opposing faces of the block gauges and one side finished at a
time, as shown in Fig. 4.6.4(8). It was a process requiring highly skilled craftsmanship. Today
most block gauges are machine-lapped, and the lapping speed, pressure, and time are precisely
controlled. .
A rotary lapping machine has a set of two circular lap tables of the same diameter; the work is
placed between the lap tables and opposing parallel surfaces are lapped simultaneously. The
workpieces are laid out and their positions are switched from time to time, as shown in Fig.
4.6.7(9,10), so that a particular piece will not follow the same path on the lap table. The lapping
process if divided into four stages. Each stage uses progressively finer abrasives to remove ihe
machining margin in stages and attain the final dimensions and form. The machining margin and
abrasive grain size for each lapping stage are shown in Table 6.4.3. Chromium oxide, silicon
carbide, aluminum oxide, and diamond are some materials used as lapping abrasives. According
to one report(11), after measurements were made on steel block gauges made by six Japanese and
ten overseas manufacturers, a surface roughness R max of 20-40 nm was found to be typical.
4.7.1 Introduction
Rolling bearings are considered to possess the highest precision among mechanical structural
parts. Rolling bearings are produced and used in the largest quantities worldwide, despite the
many types of bearings available, including sliding bearings and magnetic bearings. However,
this does not necessarily mean that rolling bearings are superior in performance to other types of
bearings. For example, the runout accuracy of rolling bearings, when mounted on a spindle, is
actually inferior to that of properly designed fluid bearings.
In terms of the accuracy of the component parts, however, rolling bearings are superior to fluid
bearings. In a fluid bearing the averaging effect of a fluid film does not allow the geometrical
inaccuracies of its component parts to exert a direct influence upon the accuracy of the spindle.
In contrast, the accuracy of the component parts of a rolling bearing directly influences the
overall accuracy of the bearing. Rolling elements such as balls and rollers, in particular, can
greatly affect the overall bearing accuracy, and therefore must be of a very high grade of
accuracy.
This section discusses ultra-precision balls for ball bearings and includes a description of the
application background that requires such ultra-precision. It also describes the manufacturing
processes and the principle followed in the processes used to produce ultra-precision balls.
Table 4.7.1 Machining margin and abrasive grain size in lapping (figures in μm)
Stage Hand lapping Machine lapping Grain diameter
Finish grinding 20 15-20 -
1st stage rough lapping 5 1.1-2.0 4-8
2nd stage rough lapping 3 0.4-0.6 2-4
1st stage finish lapping 1 0-0.2 0-1
2nd stage finish lapping Nominal dimension Nominal dimension 0-0.5
Fig. 4.7.1. Machine lapping, (a) Workpiece layout and position switch (28 pieces), lst-stage
rough lapping, multiple carriers, (b) Workpiece layout (25 pieces), lst- stage rough lapping,
multiple carriers, (c) Workpiece layout and position switch (16 pieces), 2nd-stage rough lapping
and after, single carrier.
4.8 CCDs (charge-coupled devices)
The CCD is currently used in two applications. One is the linear sensor in facsimile equipment.
The other is the area sensor for camera fields. Home video cameras are common, and in industry,
video cameras are used for quantity inspection and as robot eyes, in broadcasting for TV
cameras, and in the medical field as very small cameras for stomach diagnosis. In the near future,
there may be applications in visual telephones and as electric still cameras. Figure 4.8.1 shows a
1/3-inch (8.47 mm) 320 000 (320k)-pixel interline CCD (PAL format) made by Toshiba; Table
4.8.1 shows its main specifications.
Figure 4.8.2 shows the trend for CCDs estimated by the author, compared with the general trend
for DRAMs. In general, the main specifications for CCDs are the diagonal length (i.e. optical
format, e.g. 1/3 inch) and the (lumber of pixels (e.g. 320 000). Home video cameras have
become smaller and lighter year by year, accompanying the decrease in the size of CCDs.
Currently among MOS devices, DRAMs are at the forefront of technology. CCDs have design
rules similar to those for main-generation DRAMs, but in mass production, CCDs have about a
one-year lag. For example, the 4M generation of DRAMs corresponds to the 1/3-inch 270 000-
350 000-pixel CCDs. In CCDs, the size has decreased from 1 /2-inch (12.7 mm) to 1/3-inch and
a 1/4- inch a (6.35 mm) size will soon appear. The successful production of 1/5-inch (5.08 mm)
CCDs hinges on improving their sensitivity. On the other hand, the number of pixels has been
between 270 000 and 400 000. For high- definition TV, CCDs with 2 million pixels will be used.
1
Fig. 4.8.1. External shape and cross section of CCD. ( inch 320000 TOSHIBA).
3
Each photodiode in a pixel stores charges (electrons) in proportion to the light intensity and
transfers these charges to the VCCD (see Fig. 4.8.3). The VCCD then transfers these charges to
the next stage using a four- phase clock ( V 1 V 4 ). Thus the charges of an entire pixel
column are transferred to the HCCD. The HCCD transfers these charges at a very high speed to
the sense amplifier using a two-phase clock ( H 1 H 2 ). The sense amplifier amplifies these
charges to drive the outside load. All charges in the pixel columns are transferred to the HCCD
during one field period of TV scanning(1).
In the VCCD, charges are transferred as follows. Figure 4.8.7 shows the transfer timing for a
VCCD.
Table 4.8.1 Main specifications of the Toshiba TCD 525 ID
First, for the odd fields of TV scanning, charges in the photodiodes of the nth and (n+l)th pixels
are transferred to their VCCD and combined. Then, according to the four-phase clock pulses, the
combined charges are transferred to the next stage one by one. Next, for even fields, the charges
in the photodiodes of the (n-l)th and nth pixels are combined and transferred. The colour
information comprises a set of eight pixels as shown in Fig. 4.8.4. For odd fields, the two
combinations Gr + Ye and Mg + Ye are made in one row of the VCCD, and the other two
combinations Mg + Cy and Gr + Cy in the adjoining row. For even fields, combinations Ye +
Mg and Ye + Gr are made in one row, and Cy + Gr and Cy + Mg in the adjoining row. In a TV
receiver the RGB information is formed from this information. In the visual spectral region,
complementary colours are generally as follows: Ye = R + G; Cy = B + G; Mg = B + R.
4.8.2 Principle of CCDs
(a) Photodiode
When a field shift pulse is applied to the second polysilicon gate, free charges in the n+ layer of
the photodiode are swept out to the VCCD; the n + layer is thus depleted and new charges will
be stored in proportion to the light intensity during the next field period of TV scanning. When
light is focused on the photodiode, electrons in the valence band absorb light energy and are
excited to the conduction band. This creates pairs of electrons and holes. The holes are carried
out to the p+ layer of the photodiode and channel stopper (which surrounds the photodiode), and
electrons are stored in the n+ layer of the photodiode. Figure 4.8.8 shows the potential curve of
the photodiode.
(b) VCCD
It is first necessary to remove all free charges in the n-type buried channel of the VCCD. This is
done by applying a positive pulse to the top of the VCCD gate, which has a high voltage drain
(OD), as shown in Fig. 4.8.3. Thus the n-type buried channel layers of the VCCD are depleted
and will have positive charges. Figure 4.8.9 shows the potential curve of the VCCD. To transfer
charges to the next stage, a negative gate voltage is applied to set the bottom of the potential at ~
0 V. Conversely, to receive charges, the gate voltage is set equal to 0 V to set the bottom of the
potential at a positive voltage.
(c) HCCD
The HCCD is different from the VCCD in three respects: (1) the buried channels of a HCCD
consist of two types, n and n-, as shown in Fig. 4.8.10, and in their depleted state, the potentials
of both layers are different; (2) the HCCD is driven by a two-phase clock; (3) the transfer speed
is much faster than for the VCCD. To transfer charges to the next stage, phases H1 and H2
are set at 0 V and +5 V respectively. Charges stored in the n layer of H1 are transferred to the n
layer of H2 as shown in Fig. 4.8.10.
(d) Sense amplifier
The sense amplifier is constructed of two-stage source- followers, a floating capacitor, reset gate,
and an output gate, as shown in Fig. 4.8.11. The gate of the first-stage source-follower is
connected to the floating capacitor, which is the source of the reset gate. The output gate, located
between the floating capacitor
Fig. 4.8.2. Trend for CCDs estimated by the author, compared with the general trend for
DRAMs.
reduce smearing further, the insulator thickness should be made as thin as possible, as shown in
Fig. 7.4.12. The material of the first metal layer is usually Al, W, or some other heavy metal (4),
Even a 5 nm pinhole will cause an image defect. Therefore a second shield metal layer is applied
over the first metal layer. The insulators are deposited on the metal to form a smooth plane and
protect the metal.
On top of the silicon chip, a nitride film is deposited to block penetration by heavy metals or
alkali metals such as Na, prevent the filtration of humidity, and reduce the surface states of
silicon.
Table 4.8.3 summarizes the film thicknesses and their tolerances.
(b) Colour filter
In colour CCDs, one colour is assigned from yellow, cyan, magenta, and green on each pixel. On
top of this, a microlens is formed. The colour filter is formed directly on the silicon chip. First,
the base layer is coated to make the surface smooth. Next, a dyeing layer, usually of casein, is
coated and engraved by a photo-engraving process. The residual part is dyed by dipping in a
dyeing liquid. Then the interlayer is coated, separate between the upper and lower dyed layers.
To reduce the thickness of the colour filter, the interlayer can be omitted by special treatment of
the dyed layers, i.e. fixed colour treatment.
After three alternate layers of dyed layer and interlayer are repeated, the top layer is coated for
protection and a microlens is formed, the thickness and shape of which are controlled to focus
light on the photodiode. Green colour is produced by overlaying yellow and cyan. The total
thickness of the colour filter is ~ 6-9 μm.
Fig. 4.8.5. Cross section of unit cell.
(c) Packaging
After the colour filter is formed, a first test is made to eliminate inferior chips. Only good chips
are packaged. Figure 4.8.13 shows the packaging process, which consists of dicing, mounting,
wire-bonding, glass-lid setting, and sealing. During the process, if a 1 μm particle is deposited on
the photodiode, the chip’s performance will suffer. So the packaging process has to be carefully
conducted to prevent foreign particles on the chips. In particular, silicon chippings produced by
the dicing process are carefully removed.
(d) Testing
Two tests are performed for CCDs. The first is conducted at high temperatures to detect defects
such
Fig. 4.8.6. Basic construction of VCCD.
Fig. 4.8.12. Relation between smear ratio and thickness of 1st insulator.
are transferred among different pixel combinations. Therefore if the signals have different
magnitude, this appears as flicker in. colour. The countermeasure is to reduce variation among
pixels in the colour spectrum. 4. Fine flaws. A l00nm flaw can be apparent to the eye. Handling
is therefore carefully done in all processes, especially in packaging. These defects are checked by
the highly sensitive human eye. In other non-image devices, these flaws create no problems.
(c) Technology for achieving accuracy in aligning and working elements in the wafer process.
1. Accuracy in aligning and working. Table 4.8.4 shows the author’s estimated targets for
accuracy of aligning and working elements in CCDs, compared with the general targets for
DRAMs. The design rule for CCDs is nearly the same as for DRAMs. However, in DRAMs the
minimum working dimension is the gate length of a unit cell, whereas in CCDs it is the etched
length of the HCCD gate, indicated by L in Fig. 4.8.10. Moreover, the separation between unit
cells is achieved in a different manner. In DRAMs, separation between unit cells is usually
effected by insulators such as LOCOS (local oxidation of silicon) and the implanted impurities
underneath. In CCDs however, it is effected only by implanted impurities called channel
stoppers. This is because in CCDs there are no wirings over the gates, crossing the separations.
Signal separation between adjoining rows of VCCDs is affected by alignment errors in the
implantation process in these adjoining cells.
As shown in Fig. 4.8.16, signal charges in the
Fig. 7.4.13. Process flow of packaging.
photodiode can pour into three different ports when errors in the alignment and distributed
concentration of the implanted impurities exceed certain threshholds. The first is the port to its
own buried channel, when the shift pulse is applied to the gate. The second is the port to the
overflow drain, when an intense light impinges on the photodiode. The third is the port to the
adjoining buried channel, which when taken will cause a defect. This can be caused by an
alignment error of > 300 nm in the first P-well.
2. Angle of ion beam in the implantation process. The angle of the ion beam when implanting the
n+ and p + layers of the photodiode has a large effect on the
Fig. 4.8.14. Relation between number of input electrons to SA and output voltage, in which main
defects are plotted.
Table 7.4.4 Estimated targets for design rule and accuracy of aligning and working CCD
elements, compared with those generally used for DRAMs.
CCD l/2/270k l/3/270k l/4/270k l/5/270k 1/5/3
1/2/350k l/3/350k 1/4/350k 50k
A video cassette recorder (VCR) is a device used for recording and replaying video and audio
signals. These signals are recorded on and read out from magnetic tapes using magnetic heads. In
a VCR system, magnetic tape runs on a tape guide and a cylinder unit with magnetic heads
rotates to read signals from and write signals on to tapes. Key VCR mechanisms are the tape
running guideway and cylinder unit. As shown in Fig. 4.1 .la,/there is a very narrow gap (0.5
μm) at the top of the magnetic head. In replaying, this narrow gap must follow the track very
precisely. The setting accuracy for heads on the cylinder unit is therefore a very important factor
in picture quality and compatibility.
On conventional production lines, these magnetic heads are set manually by skilled operators, by
observing optically enlarged head images on monitor displays. In this adjustment work, tjie
position of the narrow gap is used for reference. This work is tedious, slow, and unreliable. We
have therefore developed an automatic adjustment system for VCR magnetic heads on the
cylinder unit.
calculated by an image-processing computer using two head images (with and without
interferometric fringes).
As shown in Fig. 4.9.3, two head images through two optical systems are transferred to image
processing computers and calculated parameters are transferred to the adjustment mechanism.
As shown in Figs 4.9.4 and 4.9.5 the concepts for calculating head positioning errors involve the
following. 1
1. The setting angle error is calculated from the mutual positions of the two head gap images
relative to cursor positions on the monitor display. These cursor positions are calibrated as 180°
opposite each other using the standard master cylinder unit.
Fig. 4.9.1. VCR head assembly, (a) Magnetic head, (b) Cylinder unit.
2. Rotational error for the screw axis is calculated from the distance between the centre of
the interfero-metric fringes and the centre of the gap.
3. The protrusion distance from the cylinder edge is adjusted, where the fringe contrast is
higher than a certain threshold level, using interferometric optics. The interferometric optics are
also calibrated using the master cylinder unit.
Calculation accuracies for the gap position and fringe centre directly affect the adjustment
accuracies. High-speed and highly accurate and reliable algorithms are therefore required, even
when a head image is noisy or unclear because of dust particles on the head gap.
4.9.2 Image processing algorithm
The software configuration of this system is as follows:
(1) calibration for optical magnifying force and cursor positions;
(2) autofocusing for the optical system;
1.
2. Fig. 4.10.1. Optical fibre structures.
3.
Fig. 4.10.2. Optical fibre preform fabrication by VAD.
Optical fibres can be classified into (1) step-index multimode fibres, (2) graded-index multimode
fibres, and (3) single-mode fibres, as shown in Fig. 7.9.1. The fibre diameter is typically 125 μm.
The relative refractive-index difference between core and cladding in Fig. 7.9.1 is typically
0.3-1%, depending on the fibre structure. Step-index and graded-index multimode fibres, which
have larger core diameters, several tens of micrometres, find use in short-haul transmission
systems. Single-mode fibres with smaller core diameters, ~ 10 μm, are the most widely used both
in long-haul systems and in subscriber systems, because of their low transmission losses and
high transmission bandwidths. In comparison with multimode fibres, much higher precision is
required in handling single-mode fibres and related optical components.
The first step in the fabrication of silica-based optical fibres is to compose a fibre ‘preform’
which has a similar cross-sectional geometry to the fibre product but with much larger outer
diameter of 20-50 mm and a shorter length of 30-100 cm. Three popular methods of preform
fabrication are (1) modified vapour-phase deposition (MOCVD), (2) outside vapour-phase
deposition (OVPD), and (3) vapour-phase axial deposition (VAD) (2). A VAD set-up is shown in
Fig. 7.9.2. Fine glass particles, synthesized by flame hydrolysis of SiCl 4 and GeCl4 in an oxy-
hydrogen torch, are deposited in an axial direction on the end of a rotating fused silica target rod.
The porous glass preform consisting of SiO2~GeO2 particle core and SiO2 particle cladding is
then heated to ~ 1450°C in an electric furnace for consolidation. The relative refractive-index
difference A between core and cladding is adjusted to the desired value by controlling the GeCl4
flow rate in the core torch during deposition. The fibre preform thus fabricated is heated to
higher temperature (~ 2000°C) in a carbon furnace and drawn into long optical fibres as
illustrated in Fig. 7.9.3. During fibre-drawing, the outer diameter of the fibre is precisely
controlled to 125 ± 0.5 μm by using a non-contact monitoring apparatus with a scanning He-Ne
laser beam. Simultaneously with drawing, the fibre is coated with polymer resin for surface
protection and easy handling. An optical fibre, several tens of kilometres long can be drawn from
a single preform. It should be stressed that the excellent transmission characteristics of silica-
based optical fibres are partly due to the vapour-phase deposition process by which ultra-high-
purity glass preforms can be synthesized, and partly to the fibre drawing process by which an
extremely smooth boundary (probably sub-nanometre roughness) between core and cladding can
be realized.